WIP

camera/depthai.py

@@ -10,6 +10,17 @@ import cv2
 logger = logging.getLogger(__name__)
 
 
+def to_tensor_result(packet):
+    return {
+        name: np.array(packet.getLayerFp16(name))
+        for name in [tensor.name for tensor in packet.getRaw().tensors]
+    }
+
+
+def to_planar(frame):
+    return frame.transpose(2, 0, 1).flatten()
+
+
 class FramePublisher:
     def __init__(self, mqtt_client: mqtt.Client, frame_topic: str, img_width: int, img_height: int):
         self._mqtt_client = mqtt_client
@@ -22,6 +33,72 @@ class FramePublisher:
         logger.info("configure pipeline")
         pipeline = dai.Pipeline()
 
+        version = "2021.2"
+        pipeline.setOpenVINOVersion(version=dai.OpenVINO.Version.VERSION_2021_2)
+
+        # colorCam = pipeline.create(dai.node.ColorCamera)
+        # colorCam.setPreviewSize(256, 256)
+        # colorCam.setVideoSize(1024, 1024)  # 4 times larger in both axis
+        # colorCam.setResolution(dai.ColorCameraProperties.SensorResolution.THE_1080_P)
+        # colorCam.setInterleaved(False)
+        # colorCam.setBoardSocket(dai.CameraBoardSocket.RGB)
+        # colorCam.setFps(10)
+        #
+        # controlIn = pipeline.create(dai.node.XLinkIn)
+        # controlIn.setStreamName('control')
+        # controlIn.out.link(colorCam.inputControl)
+        #
+        # cam_xout = pipeline.create(dai.node.XLinkOut)
+        # cam_xout.setStreamName('video')
+        # colorCam.video.link(cam_xout.input)
+
+        # ---------------------------------------
+        # 1st stage NN - text-detection
+        # ---------------------------------------
+
+        nn = pipeline.create(dai.node.NeuralNetwork)
+        nn.setBlobPath(
+            blobconverter.from_zoo(name="east_text_detection_256x256", zoo_type="depthai", shaves=6, version=version))
+        colorCam.preview.link(nn.input)
+
+        nn_xout = pipeline.create(dai.node.XLinkOut)
+        nn_xout.setStreamName('detections')
+        nn.out.link(nn_xout.input)
+
+        # ---------------------------------------
+        # 2nd stage NN - text-recognition-0012
+        # ---------------------------------------
+
+        manip = pipeline.create(dai.node.ImageManip)
+        manip.setWaitForConfigInput(True)
+
+        manip_img = pipeline.create(dai.node.XLinkIn)
+        manip_img.setStreamName('manip_img')
+        manip_img.out.link(manip.inputImage)
+
+        manip_cfg = pipeline.create(dai.node.XLinkIn)
+        manip_cfg.setStreamName('manip_cfg')
+        manip_cfg.out.link(manip.inputConfig)
+
+        manip_xout = pipeline.create(dai.node.XLinkOut)
+        manip_xout.setStreamName('manip_out')
+
+        nn2 = pipeline.create(dai.node.NeuralNetwork)
+        nn2.setBlobPath(blobconverter.from_zoo(name="text-recognition-0012", shaves=6, version=version))
+        nn2.setNumInferenceThreads(2)
+        manip.out.link(nn2.input)
+        manip.out.link(manip_xout.input)
+
+        nn2_xout = pipeline.create(dai.node.XLinkOut)
+        nn2_xout.setStreamName("recognitions")
+        nn2.out.link(nn2_xout.input)
+
+
+
+
+
+
+
         cam_rgb = pipeline.create(dai.node.ColorCamera)
         xout_rgb = pipeline.create(dai.node.XLinkOut)
 
@@ -40,6 +117,150 @@ class FramePublisher:
         return pipeline
 
     def run(self):
+
+        with dai.Device(self._pipeline) as device:
+            q_vid = device.getOutputQueue("video", 4, blocking=False)
+            # This should be set to block, but would get to some extreme queuing/latency!
+            q_det = device.getOutputQueue("detections", 4, blocking=False)
+
+            q_rec = device.getOutputQueue("recognitions", 4, blocking=True)
+
+            q_manip_img = device.getInputQueue("manip_img")
+            q_manip_cfg = device.getInputQueue("manip_cfg")
+            q_manip_out = device.getOutputQueue("manip_out", 4, blocking=False)
+
+            controlQueue = device.getInputQueue('control')
+
+            frame = None
+            cropped_stacked = None
+            rotated_rectangles = []
+            rec_pushed = 0
+            rec_received = 0
+            host_sync = HostSeqSync()
+
+            characters = '0123456789abcdefghijklmnopqrstuvwxyz#'
+            codec = CTCCodec(characters)
+
+            ctrl = dai.CameraControl()
+            ctrl.setAutoFocusMode(dai.CameraControl.AutoFocusMode.CONTINUOUS_VIDEO)
+            ctrl.setAutoFocusTrigger()
+            controlQueue.send(ctrl)
+
+            while True:
+                vid_in = q_vid.tryGet()
+                if vid_in is not None:
+                    host_sync.add_msg(vid_in)
+
+                # Multiple recognition results may be available, read until queue is empty
+                while True:
+                    in_rec = q_rec.tryGet()
+                    if in_rec is None:
+                        break
+                    rec_data = bboxes = np.array(in_rec.getFirstLayerFp16()).reshape(30, 1, 37)
+                    decoded_text = codec.decode(rec_data)[0]
+                    pos = rotated_rectangles[rec_received]
+                    print("{:2}: {:20}".format(rec_received, decoded_text),
+                          "center({:3},{:3}) size({:3},{:3}) angle{:5.1f} deg".format(
+                              int(pos[0][0]), int(pos[0][1]), pos[1][0], pos[1][1], pos[2]))
+                    # Draw the text on the right side of 'cropped_stacked' - placeholder
+                    if cropped_stacked is not None:
+                        cv2.putText(cropped_stacked, decoded_text,
+                                    (120 + 10, 32 * rec_received + 24),
+                                    cv2.FONT_HERSHEY_SIMPLEX, 0.8, (0, 255, 0), 2)
+                        cv2.imshow('cropped_stacked', cropped_stacked)
+                    rec_received += 1
+
+                if cv2.waitKey(1) == ord('q'):
+                    break
+
+                if rec_received >= rec_pushed:
+                    in_det = q_det.tryGet()
+                    if in_det is not None:
+                        frame = host_sync.get_msg(in_det.getSequenceNum()).getCvFrame().copy()
+
+                        scores, geom1, geom2 = to_tensor_result(in_det).values()
+                        scores = np.reshape(scores, (1, 1, 64, 64))
+                        geom1 = np.reshape(geom1, (1, 4, 64, 64))
+                        geom2 = np.reshape(geom2, (1, 1, 64, 64))
+
+                        bboxes, confs, angles = east.decode_predictions(scores, geom1, geom2)
+                        boxes, angles = east.non_max_suppression(np.array(bboxes), probs=confs, angles=np.array(angles))
+                        rotated_rectangles = [
+                            east.get_cv_rotated_rect(bbox, angle * -1)
+                            for (bbox, angle) in zip(boxes, angles)
+                        ]
+
+                        rec_received = 0
+                        rec_pushed = len(rotated_rectangles)
+                        if rec_pushed:
+                            print("====== Pushing for recognition, count:", rec_pushed)
+                        cropped_stacked = None
+                        for idx, rotated_rect in enumerate(rotated_rectangles):
+                            # Detections are done on 256x256 frames, we are sending back 1024x1024
+                            # That's why we multiply center and size values by 4
+                            rotated_rect[0][0] = rotated_rect[0][0] * 4
+                            rotated_rect[0][1] = rotated_rect[0][1] * 4
+                            rotated_rect[1][0] = rotated_rect[1][0] * 4
+                            rotated_rect[1][1] = rotated_rect[1][1] * 4
+
+                            # Draw detection crop area on input frame
+                            points = np.int0(cv2.boxPoints(rotated_rect))
+                            print(rotated_rect)
+                            cv2.polylines(frame, [points], isClosed=True, color=(255, 0, 0), thickness=1,
+                                          lineType=cv2.LINE_8)
+
+                            # TODO make it work taking args like in OpenCV:
+                            # rr = ((256, 256), (128, 64), 30)
+                            rr = dai.RotatedRect()
+                            rr.center.x = rotated_rect[0][0]
+                            rr.center.y = rotated_rect[0][1]
+                            rr.size.width = rotated_rect[1][0]
+                            rr.size.height = rotated_rect[1][1]
+                            rr.angle = rotated_rect[2]
+                            cfg = dai.ImageManipConfig()
+                            cfg.setCropRotatedRect(rr, False)
+                            cfg.setResize(120, 32)
+                            # Send frame and config to device
+                            if idx == 0:
+                                w, h, c = frame.shape
+                                imgFrame = dai.ImgFrame()
+                                imgFrame.setData(to_planar(frame))
+                                imgFrame.setType(dai.ImgFrame.Type.BGR888p)
+                                imgFrame.setWidth(w)
+                                imgFrame.setHeight(h)
+                                q_manip_img.send(imgFrame)
+                            else:
+                                cfg.setReusePreviousImage(True)
+                            q_manip_cfg.send(cfg)
+
+                            # Get manipulated image from the device
+                            transformed = q_manip_out.get().getCvFrame()
+
+                            rec_placeholder_img = np.zeros((32, 200, 3), np.uint8)
+                            transformed = np.hstack((transformed, rec_placeholder_img))
+                            if cropped_stacked is None:
+                                cropped_stacked = transformed
+                            else:
+                                cropped_stacked = np.vstack((cropped_stacked, transformed))
+
+                if cropped_stacked is not None:
+                    cv2.imshow('cropped_stacked', cropped_stacked)
+
+                if frame is not None:
+                    cv2.imshow('frame', frame)
+
+                key = cv2.waitKey(1)
+                if key == ord('q'):
+                    break
+                elif key == ord('t'):
+                    print("Autofocus trigger (and disable continuous)")
+                    ctrl = dai.CameraControl()
+                    ctrl.setAutoFocusMode(dai.CameraControl.AutoFocusMode.AUTO)
+                    ctrl.setAutoFocusTrigger()
+                    controlQueue.send(ctrl)
+
+
+
         # Connect to device and start pipeline
         with dai.Device(self._pipeline) as device:
             logger.info('MxId: %s', device.getDeviceInfo().getMxId())

camera/east.py

@@ -0,0 +1,232 @@
+import cv2
+import depthai
+import numpy as np
+
+_conf_threshold = 0.5
+
+
+def get_cv_rotated_rect(bbox, angle):
+    x0, y0, x1, y1 = bbox
+    width = abs(x0 - x1)
+    height = abs(y0 - y1)
+    x = x0 + width * 0.5
+    y = y0 + height * 0.5
+    return [x.tolist(), y.tolist()], [width.tolist(), height.tolist()], np.rad2deg(angle)
+
+
+def rotated_Rectangle(bbox, angle):
+    X0, Y0, X1, Y1 = bbox
+    width = abs(X0 - X1)
+    height = abs(Y0 - Y1)
+    x = int(X0 + width * 0.5)
+    y = int(Y0 + height * 0.5)
+
+    pt1_1 = (int(x + width / 2), int(y + height / 2))
+    pt2_1 = (int(x + width / 2), int(y - height / 2))
+    pt3_1 = (int(x - width / 2), int(y - height / 2))
+    pt4_1 = (int(x - width / 2), int(y + height / 2))
+
+    t = np.array([[np.cos(angle), -np.sin(angle), x - x * np.cos(angle) + y * np.sin(angle)],
+                  [np.sin(angle), np.cos(angle), y - x * np.sin(angle) - y * np.cos(angle)],
+                  [0, 0, 1]])
+
+    tmp_pt1_1 = np.array([[pt1_1[0]], [pt1_1[1]], [1]])
+    tmp_pt1_2 = np.dot(t, tmp_pt1_1)
+    pt1_2 = (int(tmp_pt1_2[0][0]), int(tmp_pt1_2[1][0]))
+
+    tmp_pt2_1 = np.array([[pt2_1[0]], [pt2_1[1]], [1]])
+    tmp_pt2_2 = np.dot(t, tmp_pt2_1)
+    pt2_2 = (int(tmp_pt2_2[0][0]), int(tmp_pt2_2[1][0]))
+
+    tmp_pt3_1 = np.array([[pt3_1[0]], [pt3_1[1]], [1]])
+    tmp_pt3_2 = np.dot(t, tmp_pt3_1)
+    pt3_2 = (int(tmp_pt3_2[0][0]), int(tmp_pt3_2[1][0]))
+
+    tmp_pt4_1 = np.array([[pt4_1[0]], [pt4_1[1]], [1]])
+    tmp_pt4_2 = np.dot(t, tmp_pt4_1)
+    pt4_2 = (int(tmp_pt4_2[0][0]), int(tmp_pt4_2[1][0]))
+
+    points = np.array([pt1_2, pt2_2, pt3_2, pt4_2])
+
+    return points
+
+
+def non_max_suppression(boxes, probs=None, angles=None, overlapThresh=0.3):
+    # if there are no boxes, return an empty list
+    if len(boxes) == 0:
+        return [], []
+
+    # if the bounding boxes are integers, convert them to floats -- this
+    # is important since we'll be doing a bunch of divisions
+    if boxes.dtype.kind == "i":
+        boxes = boxes.astype("float")
+
+    # initialize the list of picked indexes
+    pick = []
+
+    # grab the coordinates of the bounding boxes
+    x1 = boxes[:, 0]
+    y1 = boxes[:, 1]
+    x2 = boxes[:, 2]
+    y2 = boxes[:, 3]
+
+    # compute the area of the bounding boxes and grab the indexes to sort
+    # (in the case that no probabilities are provided, simply sort on the bottom-left y-coordinate)
+    area = (x2 - x1 + 1) * (y2 - y1 + 1)
+    idxs = y2
+
+    # if probabilities are provided, sort on them instead
+    if probs is not None:
+        idxs = probs
+
+    # sort the indexes
+    idxs = np.argsort(idxs)
+
+    # keep looping while some indexes still remain in the indexes list
+    while len(idxs) > 0:
+        # grab the last index in the indexes list and add the index value to the list of picked indexes
+        last = len(idxs) - 1
+        i = idxs[last]
+        pick.append(i)
+
+        # find the largest (x, y) coordinates for the start of the bounding box and the smallest (x, y) coordinates
+        # for the end of the bounding box
+        xx1 = np.maximum(x1[i], x1[idxs[:last]])
+        yy1 = np.maximum(y1[i], y1[idxs[:last]])
+        xx2 = np.minimum(x2[i], x2[idxs[:last]])
+        yy2 = np.minimum(y2[i], y2[idxs[:last]])
+
+        # compute the width and height of the bounding box
+        w = np.maximum(0, xx2 - xx1 + 1)
+        h = np.maximum(0, yy2 - yy1 + 1)
+
+        # compute the ratio of overlap
+        overlap = (w * h) / area[idxs[:last]]
+
+        # delete all indexes from the index list that have overlap greater than the provided overlap threshold
+        idxs = np.delete(idxs, np.concatenate(([last], np.where(overlap > overlapThresh)[0])))
+
+    # return only the bounding boxes that were picked
+    return boxes[pick].astype("int"), angles[pick]
+
+
+def decode_predictions(scores, geometry1, geometry2):
+    # grab the number of rows and columns from the scores volume, then
+    # initialize our set of bounding box rectangles and corresponding
+    # confidence scores
+    (numRows, numCols) = scores.shape[2:4]
+    rects = []
+    confidences = []
+    angles = []
+
+    # loop over the number of rows
+    for y in range(0, numRows):
+        # extract the scores (probabilities), followed by the
+        # geometrical data used to derive potential bounding box
+        # coordinates that surround text
+        scoresData = scores[0, 0, y]
+        xData0 = geometry1[0, 0, y]
+        xData1 = geometry1[0, 1, y]
+        xData2 = geometry1[0, 2, y]
+        xData3 = geometry1[0, 3, y]
+        anglesData = geometry2[0, 0, y]
+
+        # loop over the number of columns
+        for x in range(0, numCols):
+            # if our score does not have sufficient probability,
+            # ignore it
+            if scoresData[x] < _conf_threshold:
+                continue
+
+            # compute the offset factor as our resulting feature
+            # maps will be 4x smaller than the input image
+            (offsetX, offsetY) = (x * 4.0, y * 4.0)
+
+            # extract the rotation angle for the prediction and
+            # then compute the sin and cosine
+            angle = anglesData[x]
+            cos = np.cos(angle)
+            sin = np.sin(angle)
+
+            # use the geometry volume to derive the width and height
+            # of the bounding box
+            h = xData0[x] + xData2[x]
+            w = xData1[x] + xData3[x]
+
+            # compute both the starting and ending (x, y)-coordinates
+            # for the text prediction bounding box
+            endX = int(offsetX + (cos * xData1[x]) + (sin * xData2[x]))
+            endY = int(offsetY - (sin * xData1[x]) + (cos * xData2[x]))
+            startX = int(endX - w)
+            startY = int(endY - h)
+
+            # add the bounding box coordinates and probability score
+            # to our respective lists
+            rects.append((startX, startY, endX, endY))
+            confidences.append(scoresData[x])
+            angles.append(angle)
+
+    # return a tuple of the bounding boxes and associated confidences
+    return (rects, confidences, angles)
+
+
+def decode_east(nnet_packet, **kwargs):
+    scores = nnet_packet.get_tensor(0)
+    geometry1 = nnet_packet.get_tensor(1)
+    geometry2 = nnet_packet.get_tensor(2)
+    bboxes, confs, angles = decode_predictions(scores, geometry1, geometry2
+                                               )
+    boxes, angles = non_max_suppression(np.array(bboxes), probs=confs, angles=np.array(angles))
+    boxesangles = (boxes, angles)
+    return boxesangles
+
+
+def show_east(boxesangles, frame, **kwargs):
+    bboxes = boxesangles[0]
+    angles = boxesangles[1]
+    for ((X0, Y0, X1, Y1), angle) in zip(bboxes, angles):
+        width = abs(X0 - X1)
+        height = abs(Y0 - Y1)
+        cX = int(X0 + width * 0.5)
+        cY = int(Y0 + height * 0.5)
+
+        rotRect = ((cX, cY), ((X1 - X0), (Y1 - Y0)), angle * (-1))
+        points = rotated_Rectangle(frame, rotRect, color=(255, 0, 0), thickness=1)
+        cv2.polylines(frame, [points], isClosed=True, color=(255, 0, 0), thickness=1, lineType=cv2.LINE_8)
+
+    return frame
+
+
+def order_points(pts):
+    rect = np.zeros((4, 2), dtype="float32")
+    s = pts.sum(axis=1)
+    rect[0] = pts[np.argmin(s)]
+    rect[2] = pts[np.argmax(s)]
+    diff = np.diff(pts, axis=1)
+    rect[1] = pts[np.argmin(diff)]
+    rect[3] = pts[np.argmax(diff)]
+    return rect
+
+
+def four_point_transform(image, pts):
+    rect = order_points(pts)
+    (tl, tr, br, bl) = rect
+
+    widthA = np.sqrt(((br[0] - bl[0]) ** 2) + ((br[1] - bl[1]) ** 2))
+    widthB = np.sqrt(((tr[0] - tl[0]) ** 2) + ((tr[1] - tl[1]) ** 2))
+    maxWidth = max(int(widthA), int(widthB))
+
+    heightA = np.sqrt(((tr[0] - br[0]) ** 2) + ((tr[1] - br[1]) ** 2))
+    heightB = np.sqrt(((tl[0] - bl[0]) ** 2) + ((tl[1] - bl[1]) ** 2))
+    maxHeight = max(int(heightA), int(heightB))
+
+    dst = np.array([
+        [0, 0],
+        [maxWidth - 1, 0],
+        [maxWidth - 1, maxHeight - 1],
+        [0, maxHeight - 1]], dtype="float32")
+
+    M = cv2.getPerspectiveTransform(rect, dst)
+    warped = cv2.warpPerspective(image, M, (maxWidth, maxHeight))
+
+    return warped

camera/text.py

@@ -0,0 +1,61 @@
+
+class HostSeqSync:
+    def __init__(self):
+        self.imfFrames = []
+
+    def add_msg(self, msg):
+        self.imfFrames.append(msg)
+
+    def get_msg(self, target_seq):
+        for i, imgFrame in enumerate(self.imfFrames):
+            if target_seq == imgFrame.getSequenceNum():
+                self.imfFrames = self.imfFrames[i:]
+                break
+        return self.imfFrames[0]
+
+
+class CTCCodec(object):
+    """ Convert between text-label and text-index """
+
+    def __init__(self, characters):
+        # characters (str): set of the possible characters.
+        dict_character = list(characters)
+
+        self.dict = {}
+        for i, char in enumerate(dict_character):
+            self.dict[char] = i + 1
+
+        self.characters = dict_character
+        # print(self.characters)
+        # input()
+
+    def decode(self, preds):
+        """ convert text-index into text-label. """
+        texts = []
+        index = 0
+        # Select max probabilty (greedy decoding) then decode index to character
+        preds = preds.astype(np.float16)
+        preds_index = np.argmax(preds, 2)
+        preds_index = preds_index.transpose(1, 0)
+        preds_index_reshape = preds_index.reshape(-1)
+        preds_sizes = np.array([preds_index.shape[1]] * preds_index.shape[0])
+
+        for l in preds_sizes:
+            t = preds_index_reshape[index:index + l]
+
+            # NOTE: t might be zero size
+            if t.shape[0] == 0:
+                continue
+
+            char_list = []
+            for i in range(l):
+                # removing repeated characters and blank.
+                if not (i > 0 and t[i - 1] == t[i]):
+                    if self.characters[t[i]] != '#':
+                        char_list.append(self.characters[t[i]])
+            text = ''.join(char_list)
+            texts.append(text)
+
+            index += l
+
+        return texts

east.py

@@ -0,0 +1,229 @@
+import cv2
+import depthai
+import numpy as np
+
+_conf_threshold = 0.5
+
+def get_cv_rotated_rect(bbox, angle):
+    x0, y0, x1, y1 = bbox
+    width = abs(x0 - x1)
+    height = abs(y0 - y1)
+    x = x0 + width * 0.5
+    y = y0 + height * 0.5
+    return ([x.tolist(), y.tolist()], [width.tolist(), height.tolist()], np.rad2deg(angle))
+
+def rotated_Rectangle(bbox, angle):
+    X0, Y0, X1, Y1 = bbox
+    width = abs(X0 - X1)
+    height = abs(Y0 - Y1)
+    x = int(X0 + width * 0.5)
+    y = int(Y0 + height * 0.5)
+
+    pt1_1 = (int(x + width / 2), int(y + height / 2))
+    pt2_1 = (int(x + width / 2), int(y - height / 2))
+    pt3_1 = (int(x - width / 2), int(y - height / 2))
+    pt4_1 = (int(x - width / 2), int(y + height / 2))
+
+    t = np.array([[np.cos(angle), -np.sin(angle), x - x * np.cos(angle) + y * np.sin(angle)],
+                  [np.sin(angle), np.cos(angle), y - x * np.sin(angle) - y * np.cos(angle)],
+                  [0, 0, 1]])
+
+    tmp_pt1_1 = np.array([[pt1_1[0]], [pt1_1[1]], [1]])
+    tmp_pt1_2 = np.dot(t, tmp_pt1_1)
+    pt1_2 = (int(tmp_pt1_2[0][0]), int(tmp_pt1_2[1][0]))
+
+    tmp_pt2_1 = np.array([[pt2_1[0]], [pt2_1[1]], [1]])
+    tmp_pt2_2 = np.dot(t, tmp_pt2_1)
+    pt2_2 = (int(tmp_pt2_2[0][0]), int(tmp_pt2_2[1][0]))
+
+    tmp_pt3_1 = np.array([[pt3_1[0]], [pt3_1[1]], [1]])
+    tmp_pt3_2 = np.dot(t, tmp_pt3_1)
+    pt3_2 = (int(tmp_pt3_2[0][0]), int(tmp_pt3_2[1][0]))
+
+    tmp_pt4_1 = np.array([[pt4_1[0]], [pt4_1[1]], [1]])
+    tmp_pt4_2 = np.dot(t, tmp_pt4_1)
+    pt4_2 = (int(tmp_pt4_2[0][0]), int(tmp_pt4_2[1][0]))
+
+    points = np.array([pt1_2, pt2_2, pt3_2, pt4_2])
+
+    return points
+
+
+def non_max_suppression(boxes, probs=None, angles=None, overlapThresh=0.3):
+    # if there are no boxes, return an empty list
+    if len(boxes) == 0:
+        return [], []
+
+    # if the bounding boxes are integers, convert them to floats -- this
+    # is important since we'll be doing a bunch of divisions
+    if boxes.dtype.kind == "i":
+        boxes = boxes.astype("float")
+
+    # initialize the list of picked indexes
+    pick = []
+
+    # grab the coordinates of the bounding boxes
+    x1 = boxes[:, 0]
+    y1 = boxes[:, 1]
+    x2 = boxes[:, 2]
+    y2 = boxes[:, 3]
+
+    # compute the area of the bounding boxes and grab the indexes to sort
+    # (in the case that no probabilities are provided, simply sort on the bottom-left y-coordinate)
+    area = (x2 - x1 + 1) * (y2 - y1 + 1)
+    idxs = y2
+
+    # if probabilities are provided, sort on them instead
+    if probs is not None:
+        idxs = probs
+
+    # sort the indexes
+    idxs = np.argsort(idxs)
+
+    # keep looping while some indexes still remain in the indexes list
+    while len(idxs) > 0:
+        # grab the last index in the indexes list and add the index value to the list of picked indexes
+        last = len(idxs) - 1
+        i = idxs[last]
+        pick.append(i)
+
+        # find the largest (x, y) coordinates for the start of the bounding box and the smallest (x, y) coordinates for the end of the bounding box
+        xx1 = np.maximum(x1[i], x1[idxs[:last]])
+        yy1 = np.maximum(y1[i], y1[idxs[:last]])
+        xx2 = np.minimum(x2[i], x2[idxs[:last]])
+        yy2 = np.minimum(y2[i], y2[idxs[:last]])
+
+        # compute the width and height of the bounding box
+        w = np.maximum(0, xx2 - xx1 + 1)
+        h = np.maximum(0, yy2 - yy1 + 1)
+
+        # compute the ratio of overlap
+        overlap = (w * h) / area[idxs[:last]]
+
+        # delete all indexes from the index list that have overlap greater than the provided overlap threshold
+        idxs = np.delete(idxs, np.concatenate(([last], np.where(overlap > overlapThresh)[0])))
+
+    # return only the bounding boxes that were picked
+    return boxes[pick].astype("int"), angles[pick]
+
+
+def decode_predictions(scores, geometry1, geometry2):
+    # grab the number of rows and columns from the scores volume, then
+    # initialize our set of bounding box rectangles and corresponding
+    # confidence scores
+    (numRows, numCols) = scores.shape[2:4]
+    rects = []
+    confidences = []
+    angles = []
+
+    # loop over the number of rows
+    for y in range(0, numRows):
+        # extract the scores (probabilities), followed by the
+        # geometrical data used to derive potential bounding box
+        # coordinates that surround text
+        scoresData = scores[0, 0, y]
+        xData0 = geometry1[0, 0, y]
+        xData1 = geometry1[0, 1, y]
+        xData2 = geometry1[0, 2, y]
+        xData3 = geometry1[0, 3, y]
+        anglesData = geometry2[0, 0, y]
+
+        # loop over the number of columns
+        for x in range(0, numCols):
+            # if our score does not have sufficient probability,
+            # ignore it
+            if scoresData[x] < _conf_threshold:
+                continue
+
+            # compute the offset factor as our resulting feature
+            # maps will be 4x smaller than the input image
+            (offsetX, offsetY) = (x * 4.0, y * 4.0)
+
+            # extract the rotation angle for the prediction and
+            # then compute the sin and cosine
+            angle = anglesData[x]
+            cos = np.cos(angle)
+            sin = np.sin(angle)
+
+            # use the geometry volume to derive the width and height
+            # of the bounding box
+            h = xData0[x] + xData2[x]
+            w = xData1[x] + xData3[x]
+
+            # compute both the starting and ending (x, y)-coordinates
+            # for the text prediction bounding box
+            endX = int(offsetX + (cos * xData1[x]) + (sin * xData2[x]))
+            endY = int(offsetY - (sin * xData1[x]) + (cos * xData2[x]))
+            startX = int(endX - w)
+            startY = int(endY - h)
+
+            # add the bounding box coordinates and probability score
+            # to our respective lists
+            rects.append((startX, startY, endX, endY))
+            confidences.append(scoresData[x])
+            angles.append(angle)
+
+    # return a tuple of the bounding boxes and associated confidences
+    return (rects, confidences, angles)
+
+
+def decode_east(nnet_packet, **kwargs):
+    scores = nnet_packet.get_tensor(0)
+    geometry1 = nnet_packet.get_tensor(1)
+    geometry2 = nnet_packet.get_tensor(2)
+    bboxes, confs, angles = decode_predictions(scores, geometry1, geometry2
+                                               )
+    boxes, angles = non_max_suppression(np.array(bboxes), probs=confs, angles=np.array(angles))
+    boxesangles = (boxes, angles)
+    return boxesangles
+
+
+def show_east(boxesangles, frame, **kwargs):
+    bboxes = boxesangles[0]
+    angles = boxesangles[1]
+    for ((X0, Y0, X1, Y1), angle) in zip(bboxes, angles):
+        width = abs(X0 - X1)
+        height = abs(Y0 - Y1)
+        cX = int(X0 + width * 0.5)
+        cY = int(Y0 + height * 0.5)
+
+        rotRect = ((cX, cY), ((X1 - X0), (Y1 - Y0)), angle * (-1))
+        points = rotated_Rectangle(frame, rotRect, color=(255, 0, 0), thickness=1)
+        cv2.polylines(frame, [points], isClosed=True, color=(255, 0, 0), thickness=1, lineType=cv2.LINE_8)
+
+    return frame
+
+
+def order_points(pts):
+    rect = np.zeros((4, 2), dtype="float32")
+    s = pts.sum(axis=1)
+    rect[0] = pts[np.argmin(s)]
+    rect[2] = pts[np.argmax(s)]
+    diff = np.diff(pts, axis=1)
+    rect[1] = pts[np.argmin(diff)]
+    rect[3] = pts[np.argmax(diff)]
+    return rect
+
+
+def four_point_transform(image, pts):
+    rect = order_points(pts)
+    (tl, tr, br, bl) = rect
+
+    widthA = np.sqrt(((br[0] - bl[0]) ** 2) + ((br[1] - bl[1]) ** 2))
+    widthB = np.sqrt(((tr[0] - tl[0]) ** 2) + ((tr[1] - tl[1]) ** 2))
+    maxWidth = max(int(widthA), int(widthB))
+
+    heightA = np.sqrt(((tr[0] - br[0]) ** 2) + ((tr[1] - br[1]) ** 2))
+    heightB = np.sqrt(((tl[0] - bl[0]) ** 2) + ((tl[1] - bl[1]) ** 2))
+    maxHeight = max(int(heightA), int(heightB))
+
+    dst = np.array([
+        [0, 0],
+        [maxWidth - 1, 0],
+        [maxWidth - 1, maxHeight - 1],
+        [0, maxHeight - 1]], dtype="float32")
+
+    M = cv2.getPerspectiveTransform(rect, dst)
+    warped = cv2.warpPerspective(image, M, (maxWidth, maxHeight))
+
+    return warped

events/events_pb2.py

@@ -1,11 +1,10 @@
 # -*- coding: utf-8 -*-
 # Generated by the protocol buffer compiler.  DO NOT EDIT!
 # source: events/events.proto
-
-from google.protobuf.internal import enum_type_wrapper
+"""Generated protocol buffer code."""
+from google.protobuf.internal import builder as _builder
 from google.protobuf import descriptor as _descriptor
-from google.protobuf import message as _message
-from google.protobuf import reflection as _reflection
+from google.protobuf import descriptor_pool as _descriptor_pool
 from google.protobuf import symbol_database as _symbol_database
 # @@protoc_insertion_point(imports)
 
@@ -15,744 +14,40 @@ _sym_db = _symbol_database.Default()
 from google.protobuf import timestamp_pb2 as google_dot_protobuf_dot_timestamp__pb2
 
 
-DESCRIPTOR = _descriptor.FileDescriptor(
-  name='events/events.proto',
-  package='robocar.events',
-  syntax='proto3',
-  serialized_options=b'Z\006events',
-  create_key=_descriptor._internal_create_key,
-  serialized_pb=b'\n\x13\x65vents/events.proto\x12\x0erobocar.events\x1a\x1fgoogle/protobuf/timestamp.proto\"T\n\x08\x46rameRef\x12\x0c\n\x04name\x18\x01 \x01(\t\x12\n\n\x02id\x18\x02 \x01(\t\x12.\n\ncreated_at\x18\x03 \x01(\x0b\x32\x1a.google.protobuf.Timestamp\"C\n\x0c\x46rameMessage\x12$\n\x02id\x18\x01 \x01(\x0b\x32\x18.robocar.events.FrameRef\x12\r\n\x05\x66rame\x18\x02 \x01(\x0c\"d\n\x0fSteeringMessage\x12\x10\n\x08steering\x18\x01 \x01(\x02\x12\x12\n\nconfidence\x18\x02 \x01(\x02\x12+\n\tframe_ref\x18\x03 \x01(\x0b\x32\x18.robocar.events.FrameRef\"d\n\x0fThrottleMessage\x12\x10\n\x08throttle\x18\x01 \x01(\x02\x12\x12\n\nconfidence\x18\x02 \x01(\x02\x12+\n\tframe_ref\x18\x03 \x01(\x0b\x32\x18.robocar.events.FrameRef\"A\n\x10\x44riveModeMessage\x12-\n\ndrive_mode\x18\x01 \x01(\x0e\x32\x19.robocar.events.DriveMode\"f\n\x0eObjectsMessage\x12\'\n\x07objects\x18\x01 \x03(\x0b\x32\x16.robocar.events.Object\x12+\n\tframe_ref\x18\x02 \x01(\x0b\x32\x18.robocar.events.FrameRef\"\x80\x01\n\x06Object\x12(\n\x04type\x18\x01 \x01(\x0e\x32\x1a.robocar.events.TypeObject\x12\x0c\n\x04left\x18\x02 \x01(\x05\x12\x0b\n\x03top\x18\x03 \x01(\x05\x12\r\n\x05right\x18\x04 \x01(\x05\x12\x0e\n\x06\x62ottom\x18\x05 \x01(\x05\x12\x12\n\nconfidence\x18\x06 \x01(\x02\"&\n\x13SwitchRecordMessage\x12\x0f\n\x07\x65nabled\x18\x01 \x01(\x08\"\x8c\x01\n\x0bRoadMessage\x12&\n\x07\x63ontour\x18\x01 \x03(\x0b\x32\x15.robocar.events.Point\x12(\n\x07\x65llipse\x18\x02 \x01(\x0b\x32\x17.robocar.events.Ellipse\x12+\n\tframe_ref\x18\x03 \x01(\x0b\x32\x18.robocar.events.FrameRef\"\x1d\n\x05Point\x12\t\n\x01x\x18\x01 \x01(\x05\x12\t\n\x01y\x18\x02 \x01(\x05\"r\n\x07\x45llipse\x12%\n\x06\x63\x65nter\x18\x01 \x01(\x0b\x32\x15.robocar.events.Point\x12\r\n\x05width\x18\x02 \x01(\x05\x12\x0e\n\x06height\x18\x03 \x01(\x05\x12\r\n\x05\x61ngle\x18\x04 \x01(\x02\x12\x12\n\nconfidence\x18\x05 \x01(\x02\"\x82\x01\n\rRecordMessage\x12+\n\x05\x66rame\x18\x01 \x01(\x0b\x32\x1c.robocar.events.FrameMessage\x12\x31\n\x08steering\x18\x02 \x01(\x0b\x32\x1f.robocar.events.SteeringMessage\x12\x11\n\trecordSet\x18\x03 \x01(\t*-\n\tDriveMode\x12\x0b\n\x07INVALID\x10\x00\x12\x08\n\x04USER\x10\x01\x12\t\n\x05PILOT\x10\x02*2\n\nTypeObject\x12\x07\n\x03\x41NY\x10\x00\x12\x07\n\x03\x43\x41R\x10\x01\x12\x08\n\x04\x42UMP\x10\x02\x12\x08\n\x04PLOT\x10\x03\x42\x08Z\x06\x65ventsb\x06proto3'
-  ,
-  dependencies=[google_dot_protobuf_dot_timestamp__pb2.DESCRIPTOR,])
-
-_DRIVEMODE = _descriptor.EnumDescriptor(
-  name='DriveMode',
-  full_name='robocar.events.DriveMode',
-  filename=None,
-  file=DESCRIPTOR,
-  create_key=_descriptor._internal_create_key,
-  values=[
-    _descriptor.EnumValueDescriptor(
-      name='INVALID', index=0, number=0,
-      serialized_options=None,
-      type=None,
-      create_key=_descriptor._internal_create_key),
-    _descriptor.EnumValueDescriptor(
-      name='USER', index=1, number=1,
-      serialized_options=None,
-      type=None,
-      create_key=_descriptor._internal_create_key),
-    _descriptor.EnumValueDescriptor(
-      name='PILOT', index=2, number=2,
-      serialized_options=None,
-      type=None,
-      create_key=_descriptor._internal_create_key),
-  ],
-  containing_type=None,
-  serialized_options=None,
-  serialized_start=1196,
-  serialized_end=1241,
-)
-_sym_db.RegisterEnumDescriptor(_DRIVEMODE)
-
-DriveMode = enum_type_wrapper.EnumTypeWrapper(_DRIVEMODE)
-_TYPEOBJECT = _descriptor.EnumDescriptor(
-  name='TypeObject',
-  full_name='robocar.events.TypeObject',
-  filename=None,
-  file=DESCRIPTOR,
-  create_key=_descriptor._internal_create_key,
-  values=[
-    _descriptor.EnumValueDescriptor(
-      name='ANY', index=0, number=0,
-      serialized_options=None,
-      type=None,
-      create_key=_descriptor._internal_create_key),
-    _descriptor.EnumValueDescriptor(
-      name='CAR', index=1, number=1,
-      serialized_options=None,
-      type=None,
-      create_key=_descriptor._internal_create_key),
-    _descriptor.EnumValueDescriptor(
-      name='BUMP', index=2, number=2,
-      serialized_options=None,
-      type=None,
-      create_key=_descriptor._internal_create_key),
-    _descriptor.EnumValueDescriptor(
-      name='PLOT', index=3, number=3,
-      serialized_options=None,
-      type=None,
-      create_key=_descriptor._internal_create_key),
-  ],
-  containing_type=None,
-  serialized_options=None,
-  serialized_start=1243,
-  serialized_end=1293,
-)
-_sym_db.RegisterEnumDescriptor(_TYPEOBJECT)
-
-TypeObject = enum_type_wrapper.EnumTypeWrapper(_TYPEOBJECT)
-INVALID = 0
-USER = 1
-PILOT = 2
-ANY = 0
-CAR = 1
-BUMP = 2
-PLOT = 3
-
-
-
-_FRAMEREF = _descriptor.Descriptor(
-  name='FrameRef',
-  full_name='robocar.events.FrameRef',
-  filename=None,
-  file=DESCRIPTOR,
-  containing_type=None,
-  create_key=_descriptor._internal_create_key,
-  fields=[
-    _descriptor.FieldDescriptor(
-      name='name', full_name='robocar.events.FrameRef.name', index=0,
-      number=1, type=9, cpp_type=9, label=1,
-      has_default_value=False, default_value=b"".decode('utf-8'),
-      message_type=None, enum_type=None, containing_type=None,
-      is_extension=False, extension_scope=None,
-      serialized_options=None, file=DESCRIPTOR,  create_key=_descriptor._internal_create_key),
-    _descriptor.FieldDescriptor(
-      name='id', full_name='robocar.events.FrameRef.id', index=1,
-      number=2, type=9, cpp_type=9, label=1,
-      has_default_value=False, default_value=b"".decode('utf-8'),
-      message_type=None, enum_type=None, containing_type=None,
-      is_extension=False, extension_scope=None,
-      serialized_options=None, file=DESCRIPTOR,  create_key=_descriptor._internal_create_key),
-    _descriptor.FieldDescriptor(
-      name='created_at', full_name='robocar.events.FrameRef.created_at', index=2,
-      number=3, type=11, cpp_type=10, label=1,
-      has_default_value=False, default_value=None,
-      message_type=None, enum_type=None, containing_type=None,
-      is_extension=False, extension_scope=None,
-      serialized_options=None, file=DESCRIPTOR,  create_key=_descriptor._internal_create_key),
-  ],
-  extensions=[
-  ],
-  nested_types=[],
-  enum_types=[
-  ],
-  serialized_options=None,
-  is_extendable=False,
-  syntax='proto3',
-  extension_ranges=[],
-  oneofs=[
-  ],
-  serialized_start=72,
-  serialized_end=156,
-)
-
-
-_FRAMEMESSAGE = _descriptor.Descriptor(
-  name='FrameMessage',
-  full_name='robocar.events.FrameMessage',
-  filename=None,
-  file=DESCRIPTOR,
-  containing_type=None,
-  create_key=_descriptor._internal_create_key,
-  fields=[
-    _descriptor.FieldDescriptor(
-      name='id', full_name='robocar.events.FrameMessage.id', index=0,
-      number=1, type=11, cpp_type=10, label=1,
-      has_default_value=False, default_value=None,
-      message_type=None, enum_type=None, containing_type=None,
-      is_extension=False, extension_scope=None,
-      serialized_options=None, file=DESCRIPTOR,  create_key=_descriptor._internal_create_key),
-    _descriptor.FieldDescriptor(
-      name='frame', full_name='robocar.events.FrameMessage.frame', index=1,
-      number=2, type=12, cpp_type=9, label=1,
-      has_default_value=False, default_value=b"",
-      message_type=None, enum_type=None, containing_type=None,
-      is_extension=False, extension_scope=None,
-      serialized_options=None, file=DESCRIPTOR,  create_key=_descriptor._internal_create_key),
-  ],
-  extensions=[
-  ],
-  nested_types=[],
-  enum_types=[
-  ],
-  serialized_options=None,
-  is_extendable=False,
-  syntax='proto3',
-  extension_ranges=[],
-  oneofs=[
-  ],
-  serialized_start=158,
-  serialized_end=225,
-)
-
-
-_STEERINGMESSAGE = _descriptor.Descriptor(
-  name='SteeringMessage',
-  full_name='robocar.events.SteeringMessage',
-  filename=None,
-  file=DESCRIPTOR,
-  containing_type=None,
-  create_key=_descriptor._internal_create_key,
-  fields=[
-    _descriptor.FieldDescriptor(
-      name='steering', full_name='robocar.events.SteeringMessage.steering', index=0,
-      number=1, type=2, cpp_type=6, label=1,
-      has_default_value=False, default_value=float(0),
-      message_type=None, enum_type=None, containing_type=None,
-      is_extension=False, extension_scope=None,
-      serialized_options=None, file=DESCRIPTOR,  create_key=_descriptor._internal_create_key),
-    _descriptor.FieldDescriptor(
-      name='confidence', full_name='robocar.events.SteeringMessage.confidence', index=1,
-      number=2, type=2, cpp_type=6, label=1,
-      has_default_value=False, default_value=float(0),
-      message_type=None, enum_type=None, containing_type=None,
-      is_extension=False, extension_scope=None,
-      serialized_options=None, file=DESCRIPTOR,  create_key=_descriptor._internal_create_key),
-    _descriptor.FieldDescriptor(
-      name='frame_ref', full_name='robocar.events.SteeringMessage.frame_ref', index=2,
-      number=3, type=11, cpp_type=10, label=1,
-      has_default_value=False, default_value=None,
-      message_type=None, enum_type=None, containing_type=None,
-      is_extension=False, extension_scope=None,
-      serialized_options=None, file=DESCRIPTOR,  create_key=_descriptor._internal_create_key),
-  ],
-  extensions=[
-  ],
-  nested_types=[],
-  enum_types=[
-  ],
-  serialized_options=None,
-  is_extendable=False,
-  syntax='proto3',
-  extension_ranges=[],
-  oneofs=[
-  ],
-  serialized_start=227,
-  serialized_end=327,
-)
-
-
-_THROTTLEMESSAGE = _descriptor.Descriptor(
-  name='ThrottleMessage',
-  full_name='robocar.events.ThrottleMessage',
-  filename=None,
-  file=DESCRIPTOR,
-  containing_type=None,
-  create_key=_descriptor._internal_create_key,
-  fields=[
-    _descriptor.FieldDescriptor(
-      name='throttle', full_name='robocar.events.ThrottleMessage.throttle', index=0,
-      number=1, type=2, cpp_type=6, label=1,
-      has_default_value=False, default_value=float(0),
-      message_type=None, enum_type=None, containing_type=None,
-      is_extension=False, extension_scope=None,
-      serialized_options=None, file=DESCRIPTOR,  create_key=_descriptor._internal_create_key),
-    _descriptor.FieldDescriptor(
-      name='confidence', full_name='robocar.events.ThrottleMessage.confidence', index=1,
-      number=2, type=2, cpp_type=6, label=1,
-      has_default_value=False, default_value=float(0),
-      message_type=None, enum_type=None, containing_type=None,
-      is_extension=False, extension_scope=None,
-      serialized_options=None, file=DESCRIPTOR,  create_key=_descriptor._internal_create_key),
-    _descriptor.FieldDescriptor(
-      name='frame_ref', full_name='robocar.events.ThrottleMessage.frame_ref', index=2,
-      number=3, type=11, cpp_type=10, label=1,
-      has_default_value=False, default_value=None,
-      message_type=None, enum_type=None, containing_type=None,
-      is_extension=False, extension_scope=None,
-      serialized_options=None, file=DESCRIPTOR,  create_key=_descriptor._internal_create_key),
-  ],
-  extensions=[
-  ],
-  nested_types=[],
-  enum_types=[
-  ],
-  serialized_options=None,
-  is_extendable=False,
-  syntax='proto3',
-  extension_ranges=[],
-  oneofs=[
-  ],
-  serialized_start=329,
-  serialized_end=429,
-)
-
-
-_DRIVEMODEMESSAGE = _descriptor.Descriptor(
-  name='DriveModeMessage',
-  full_name='robocar.events.DriveModeMessage',
-  filename=None,
-  file=DESCRIPTOR,
-  containing_type=None,
-  create_key=_descriptor._internal_create_key,
-  fields=[
-    _descriptor.FieldDescriptor(
-      name='drive_mode', full_name='robocar.events.DriveModeMessage.drive_mode', index=0,
-      number=1, type=14, cpp_type=8, label=1,
-      has_default_value=False, default_value=0,
-      message_type=None, enum_type=None, containing_type=None,
-      is_extension=False, extension_scope=None,
-      serialized_options=None, file=DESCRIPTOR,  create_key=_descriptor._internal_create_key),
-  ],
-  extensions=[
-  ],
-  nested_types=[],
-  enum_types=[
-  ],
-  serialized_options=None,
-  is_extendable=False,
-  syntax='proto3',
-  extension_ranges=[],
-  oneofs=[
-  ],
-  serialized_start=431,
-  serialized_end=496,
-)
-
-
-_OBJECTSMESSAGE = _descriptor.Descriptor(
-  name='ObjectsMessage',
-  full_name='robocar.events.ObjectsMessage',
-  filename=None,
-  file=DESCRIPTOR,
-  containing_type=None,
-  create_key=_descriptor._internal_create_key,
-  fields=[
-    _descriptor.FieldDescriptor(
-      name='objects', full_name='robocar.events.ObjectsMessage.objects', index=0,
-      number=1, type=11, cpp_type=10, label=3,
-      has_default_value=False, default_value=[],
-      message_type=None, enum_type=None, containing_type=None,
-      is_extension=False, extension_scope=None,
-      serialized_options=None, file=DESCRIPTOR,  create_key=_descriptor._internal_create_key),
-    _descriptor.FieldDescriptor(
-      name='frame_ref', full_name='robocar.events.ObjectsMessage.frame_ref', index=1,
-      number=2, type=11, cpp_type=10, label=1,
-      has_default_value=False, default_value=None,
-      message_type=None, enum_type=None, containing_type=None,
-      is_extension=False, extension_scope=None,
-      serialized_options=None, file=DESCRIPTOR,  create_key=_descriptor._internal_create_key),
-  ],
-  extensions=[
-  ],
-  nested_types=[],
-  enum_types=[
-  ],
-  serialized_options=None,
-  is_extendable=False,
-  syntax='proto3',
-  extension_ranges=[],
-  oneofs=[
-  ],
-  serialized_start=498,
-  serialized_end=600,
-)
-
-
-_OBJECT = _descriptor.Descriptor(
-  name='Object',
-  full_name='robocar.events.Object',
-  filename=None,
-  file=DESCRIPTOR,
-  containing_type=None,
-  create_key=_descriptor._internal_create_key,
-  fields=[
-    _descriptor.FieldDescriptor(
-      name='type', full_name='robocar.events.Object.type', index=0,
-      number=1, type=14, cpp_type=8, label=1,
-      has_default_value=False, default_value=0,
-      message_type=None, enum_type=None, containing_type=None,
-      is_extension=False, extension_scope=None,
-      serialized_options=None, file=DESCRIPTOR,  create_key=_descriptor._internal_create_key),
-    _descriptor.FieldDescriptor(
-      name='left', full_name='robocar.events.Object.left', index=1,
-      number=2, type=5, cpp_type=1, label=1,
-      has_default_value=False, default_value=0,
-      message_type=None, enum_type=None, containing_type=None,
-      is_extension=False, extension_scope=None,
-      serialized_options=None, file=DESCRIPTOR,  create_key=_descriptor._internal_create_key),
-    _descriptor.FieldDescriptor(
-      name='top', full_name='robocar.events.Object.top', index=2,
-      number=3, type=5, cpp_type=1, label=1,
-      has_default_value=False, default_value=0,
-      message_type=None, enum_type=None, containing_type=None,
-      is_extension=False, extension_scope=None,
-      serialized_options=None, file=DESCRIPTOR,  create_key=_descriptor._internal_create_key),
-    _descriptor.FieldDescriptor(
-      name='right', full_name='robocar.events.Object.right', index=3,
-      number=4, type=5, cpp_type=1, label=1,
-      has_default_value=False, default_value=0,
-      message_type=None, enum_type=None, containing_type=None,
-      is_extension=False, extension_scope=None,
-      serialized_options=None, file=DESCRIPTOR,  create_key=_descriptor._internal_create_key),
-    _descriptor.FieldDescriptor(
-      name='bottom', full_name='robocar.events.Object.bottom', index=4,
-      number=5, type=5, cpp_type=1, label=1,
-      has_default_value=False, default_value=0,
-      message_type=None, enum_type=None, containing_type=None,
-      is_extension=False, extension_scope=None,
-      serialized_options=None, file=DESCRIPTOR,  create_key=_descriptor._internal_create_key),
-    _descriptor.FieldDescriptor(
-      name='confidence', full_name='robocar.events.Object.confidence', index=5,
-      number=6, type=2, cpp_type=6, label=1,
-      has_default_value=False, default_value=float(0),
-      message_type=None, enum_type=None, containing_type=None,
-      is_extension=False, extension_scope=None,
-      serialized_options=None, file=DESCRIPTOR,  create_key=_descriptor._internal_create_key),
-  ],
-  extensions=[
-  ],
-  nested_types=[],
-  enum_types=[
-  ],
-  serialized_options=None,
-  is_extendable=False,
-  syntax='proto3',
-  extension_ranges=[],
-  oneofs=[
-  ],
-  serialized_start=603,
-  serialized_end=731,
-)
-
-
-_SWITCHRECORDMESSAGE = _descriptor.Descriptor(
-  name='SwitchRecordMessage',
-  full_name='robocar.events.SwitchRecordMessage',
-  filename=None,
-  file=DESCRIPTOR,
-  containing_type=None,
-  create_key=_descriptor._internal_create_key,
-  fields=[
-    _descriptor.FieldDescriptor(
-      name='enabled', full_name='robocar.events.SwitchRecordMessage.enabled', index=0,
-      number=1, type=8, cpp_type=7, label=1,
-      has_default_value=False, default_value=False,
-      message_type=None, enum_type=None, containing_type=None,
-      is_extension=False, extension_scope=None,
-      serialized_options=None, file=DESCRIPTOR,  create_key=_descriptor._internal_create_key),
-  ],
-  extensions=[
-  ],
-  nested_types=[],
-  enum_types=[
-  ],
-  serialized_options=None,
-  is_extendable=False,
-  syntax='proto3',
-  extension_ranges=[],
-  oneofs=[
-  ],
-  serialized_start=733,
-  serialized_end=771,
-)
-
-
-_ROADMESSAGE = _descriptor.Descriptor(
-  name='RoadMessage',
-  full_name='robocar.events.RoadMessage',
-  filename=None,
-  file=DESCRIPTOR,
-  containing_type=None,
-  create_key=_descriptor._internal_create_key,
-  fields=[
-    _descriptor.FieldDescriptor(
-      name='contour', full_name='robocar.events.RoadMessage.contour', index=0,
-      number=1, type=11, cpp_type=10, label=3,
-      has_default_value=False, default_value=[],
-      message_type=None, enum_type=None, containing_type=None,
-      is_extension=False, extension_scope=None,
-      serialized_options=None, file=DESCRIPTOR,  create_key=_descriptor._internal_create_key),
-    _descriptor.FieldDescriptor(
-      name='ellipse', full_name='robocar.events.RoadMessage.ellipse', index=1,
-      number=2, type=11, cpp_type=10, label=1,
-      has_default_value=False, default_value=None,
-      message_type=None, enum_type=None, containing_type=None,
-      is_extension=False, extension_scope=None,
-      serialized_options=None, file=DESCRIPTOR,  create_key=_descriptor._internal_create_key),
-    _descriptor.FieldDescriptor(
-      name='frame_ref', full_name='robocar.events.RoadMessage.frame_ref', index=2,
-      number=3, type=11, cpp_type=10, label=1,
-      has_default_value=False, default_value=None,
-      message_type=None, enum_type=None, containing_type=None,
-      is_extension=False, extension_scope=None,
-      serialized_options=None, file=DESCRIPTOR,  create_key=_descriptor._internal_create_key),
-  ],
-  extensions=[
-  ],
-  nested_types=[],
-  enum_types=[
-  ],
-  serialized_options=None,
-  is_extendable=False,
-  syntax='proto3',
-  extension_ranges=[],
-  oneofs=[
-  ],
-  serialized_start=774,
-  serialized_end=914,
-)
-
-
-_POINT = _descriptor.Descriptor(
-  name='Point',
-  full_name='robocar.events.Point',
-  filename=None,
-  file=DESCRIPTOR,
-  containing_type=None,
-  create_key=_descriptor._internal_create_key,
-  fields=[
-    _descriptor.FieldDescriptor(
-      name='x', full_name='robocar.events.Point.x', index=0,
-      number=1, type=5, cpp_type=1, label=1,
-      has_default_value=False, default_value=0,
-      message_type=None, enum_type=None, containing_type=None,
-      is_extension=False, extension_scope=None,
-      serialized_options=None, file=DESCRIPTOR,  create_key=_descriptor._internal_create_key),
-    _descriptor.FieldDescriptor(
-      name='y', full_name='robocar.events.Point.y', index=1,
-      number=2, type=5, cpp_type=1, label=1,
-      has_default_value=False, default_value=0,
-      message_type=None, enum_type=None, containing_type=None,
-      is_extension=False, extension_scope=None,
-      serialized_options=None, file=DESCRIPTOR,  create_key=_descriptor._internal_create_key),
-  ],
-  extensions=[
-  ],
-  nested_types=[],
-  enum_types=[
-  ],
-  serialized_options=None,
-  is_extendable=False,
-  syntax='proto3',
-  extension_ranges=[],
-  oneofs=[
-  ],
-  serialized_start=916,
-  serialized_end=945,
-)
-
-
-_ELLIPSE = _descriptor.Descriptor(
-  name='Ellipse',
-  full_name='robocar.events.Ellipse',
-  filename=None,
-  file=DESCRIPTOR,
-  containing_type=None,
-  create_key=_descriptor._internal_create_key,
-  fields=[
-    _descriptor.FieldDescriptor(
-      name='center', full_name='robocar.events.Ellipse.center', index=0,
-      number=1, type=11, cpp_type=10, label=1,
-      has_default_value=False, default_value=None,
-      message_type=None, enum_type=None, containing_type=None,
-      is_extension=False, extension_scope=None,
-      serialized_options=None, file=DESCRIPTOR,  create_key=_descriptor._internal_create_key),
-    _descriptor.FieldDescriptor(
-      name='width', full_name='robocar.events.Ellipse.width', index=1,
-      number=2, type=5, cpp_type=1, label=1,
-      has_default_value=False, default_value=0,
-      message_type=None, enum_type=None, containing_type=None,
-      is_extension=False, extension_scope=None,
-      serialized_options=None, file=DESCRIPTOR,  create_key=_descriptor._internal_create_key),
-    _descriptor.FieldDescriptor(
-      name='height', full_name='robocar.events.Ellipse.height', index=2,
-      number=3, type=5, cpp_type=1, label=1,
-      has_default_value=False, default_value=0,
-      message_type=None, enum_type=None, containing_type=None,
-      is_extension=False, extension_scope=None,
-      serialized_options=None, file=DESCRIPTOR,  create_key=_descriptor._internal_create_key),
-    _descriptor.FieldDescriptor(
-      name='angle', full_name='robocar.events.Ellipse.angle', index=3,
-      number=4, type=2, cpp_type=6, label=1,
-      has_default_value=False, default_value=float(0),
-      message_type=None, enum_type=None, containing_type=None,
-      is_extension=False, extension_scope=None,
-      serialized_options=None, file=DESCRIPTOR,  create_key=_descriptor._internal_create_key),
-    _descriptor.FieldDescriptor(
-      name='confidence', full_name='robocar.events.Ellipse.confidence', index=4,
-      number=5, type=2, cpp_type=6, label=1,
-      has_default_value=False, default_value=float(0),
-      message_type=None, enum_type=None, containing_type=None,
-      is_extension=False, extension_scope=None,
-      serialized_options=None, file=DESCRIPTOR,  create_key=_descriptor._internal_create_key),
-  ],
-  extensions=[
-  ],
-  nested_types=[],
-  enum_types=[
-  ],
-  serialized_options=None,
-  is_extendable=False,
-  syntax='proto3',
-  extension_ranges=[],
-  oneofs=[
-  ],
-  serialized_start=947,
-  serialized_end=1061,
-)
-
-
-_RECORDMESSAGE = _descriptor.Descriptor(
-  name='RecordMessage',
-  full_name='robocar.events.RecordMessage',
-  filename=None,
-  file=DESCRIPTOR,
-  containing_type=None,
-  create_key=_descriptor._internal_create_key,
-  fields=[
-    _descriptor.FieldDescriptor(
-      name='frame', full_name='robocar.events.RecordMessage.frame', index=0,
-      number=1, type=11, cpp_type=10, label=1,
-      has_default_value=False, default_value=None,
-      message_type=None, enum_type=None, containing_type=None,
-      is_extension=False, extension_scope=None,
-      serialized_options=None, file=DESCRIPTOR,  create_key=_descriptor._internal_create_key),
-    _descriptor.FieldDescriptor(
-      name='steering', full_name='robocar.events.RecordMessage.steering', index=1,
-      number=2, type=11, cpp_type=10, label=1,
-      has_default_value=False, default_value=None,
-      message_type=None, enum_type=None, containing_type=None,
-      is_extension=False, extension_scope=None,
-      serialized_options=None, file=DESCRIPTOR,  create_key=_descriptor._internal_create_key),
-    _descriptor.FieldDescriptor(
-      name='recordSet', full_name='robocar.events.RecordMessage.recordSet', index=2,
-      number=3, type=9, cpp_type=9, label=1,
-      has_default_value=False, default_value=b"".decode('utf-8'),
-      message_type=None, enum_type=None, containing_type=None,
-      is_extension=False, extension_scope=None,
-      serialized_options=None, file=DESCRIPTOR,  create_key=_descriptor._internal_create_key),
-  ],
-  extensions=[
-  ],
-  nested_types=[],
-  enum_types=[
-  ],
-  serialized_options=None,
-  is_extendable=False,
-  syntax='proto3',
-  extension_ranges=[],
-  oneofs=[
-  ],
-  serialized_start=1064,
-  serialized_end=1194,
-)
-
-_FRAMEREF.fields_by_name['created_at'].message_type = google_dot_protobuf_dot_timestamp__pb2._TIMESTAMP
-_FRAMEMESSAGE.fields_by_name['id'].message_type = _FRAMEREF
-_STEERINGMESSAGE.fields_by_name['frame_ref'].message_type = _FRAMEREF
-_THROTTLEMESSAGE.fields_by_name['frame_ref'].message_type = _FRAMEREF
-_DRIVEMODEMESSAGE.fields_by_name['drive_mode'].enum_type = _DRIVEMODE
-_OBJECTSMESSAGE.fields_by_name['objects'].message_type = _OBJECT
-_OBJECTSMESSAGE.fields_by_name['frame_ref'].message_type = _FRAMEREF
-_OBJECT.fields_by_name['type'].enum_type = _TYPEOBJECT
-_ROADMESSAGE.fields_by_name['contour'].message_type = _POINT
-_ROADMESSAGE.fields_by_name['ellipse'].message_type = _ELLIPSE
-_ROADMESSAGE.fields_by_name['frame_ref'].message_type = _FRAMEREF
-_ELLIPSE.fields_by_name['center'].message_type = _POINT
-_RECORDMESSAGE.fields_by_name['frame'].message_type = _FRAMEMESSAGE
-_RECORDMESSAGE.fields_by_name['steering'].message_type = _STEERINGMESSAGE
-DESCRIPTOR.message_types_by_name['FrameRef'] = _FRAMEREF
-DESCRIPTOR.message_types_by_name['FrameMessage'] = _FRAMEMESSAGE
-DESCRIPTOR.message_types_by_name['SteeringMessage'] = _STEERINGMESSAGE
-DESCRIPTOR.message_types_by_name['ThrottleMessage'] = _THROTTLEMESSAGE
-DESCRIPTOR.message_types_by_name['DriveModeMessage'] = _DRIVEMODEMESSAGE
-DESCRIPTOR.message_types_by_name['ObjectsMessage'] = _OBJECTSMESSAGE
-DESCRIPTOR.message_types_by_name['Object'] = _OBJECT
-DESCRIPTOR.message_types_by_name['SwitchRecordMessage'] = _SWITCHRECORDMESSAGE
-DESCRIPTOR.message_types_by_name['RoadMessage'] = _ROADMESSAGE
-DESCRIPTOR.message_types_by_name['Point'] = _POINT
-DESCRIPTOR.message_types_by_name['Ellipse'] = _ELLIPSE
-DESCRIPTOR.message_types_by_name['RecordMessage'] = _RECORDMESSAGE
-DESCRIPTOR.enum_types_by_name['DriveMode'] = _DRIVEMODE
-DESCRIPTOR.enum_types_by_name['TypeObject'] = _TYPEOBJECT
-_sym_db.RegisterFileDescriptor(DESCRIPTOR)
-
-FrameRef = _reflection.GeneratedProtocolMessageType('FrameRef', (_message.Message,), {
-  'DESCRIPTOR' : _FRAMEREF,
-  '__module__' : 'events.events_pb2'
-  # @@protoc_insertion_point(class_scope:robocar.events.FrameRef)
-  })
-_sym_db.RegisterMessage(FrameRef)
-
-FrameMessage = _reflection.GeneratedProtocolMessageType('FrameMessage', (_message.Message,), {
-  'DESCRIPTOR' : _FRAMEMESSAGE,
-  '__module__' : 'events.events_pb2'
-  # @@protoc_insertion_point(class_scope:robocar.events.FrameMessage)
-  })
-_sym_db.RegisterMessage(FrameMessage)
-
-SteeringMessage = _reflection.GeneratedProtocolMessageType('SteeringMessage', (_message.Message,), {
-  'DESCRIPTOR' : _STEERINGMESSAGE,
-  '__module__' : 'events.events_pb2'
-  # @@protoc_insertion_point(class_scope:robocar.events.SteeringMessage)
-  })
-_sym_db.RegisterMessage(SteeringMessage)
-
-ThrottleMessage = _reflection.GeneratedProtocolMessageType('ThrottleMessage', (_message.Message,), {
-  'DESCRIPTOR' : _THROTTLEMESSAGE,
-  '__module__' : 'events.events_pb2'
-  # @@protoc_insertion_point(class_scope:robocar.events.ThrottleMessage)
-  })
-_sym_db.RegisterMessage(ThrottleMessage)
-
-DriveModeMessage = _reflection.GeneratedProtocolMessageType('DriveModeMessage', (_message.Message,), {
-  'DESCRIPTOR' : _DRIVEMODEMESSAGE,
-  '__module__' : 'events.events_pb2'
-  # @@protoc_insertion_point(class_scope:robocar.events.DriveModeMessage)
-  })
-_sym_db.RegisterMessage(DriveModeMessage)
-
-ObjectsMessage = _reflection.GeneratedProtocolMessageType('ObjectsMessage', (_message.Message,), {
-  'DESCRIPTOR' : _OBJECTSMESSAGE,
-  '__module__' : 'events.events_pb2'
-  # @@protoc_insertion_point(class_scope:robocar.events.ObjectsMessage)
-  })
-_sym_db.RegisterMessage(ObjectsMessage)
-
-Object = _reflection.GeneratedProtocolMessageType('Object', (_message.Message,), {
-  'DESCRIPTOR' : _OBJECT,
-  '__module__' : 'events.events_pb2'
-  # @@protoc_insertion_point(class_scope:robocar.events.Object)
-  })
-_sym_db.RegisterMessage(Object)
-
-SwitchRecordMessage = _reflection.GeneratedProtocolMessageType('SwitchRecordMessage', (_message.Message,), {
-  'DESCRIPTOR' : _SWITCHRECORDMESSAGE,
-  '__module__' : 'events.events_pb2'
-  # @@protoc_insertion_point(class_scope:robocar.events.SwitchRecordMessage)
-  })
-_sym_db.RegisterMessage(SwitchRecordMessage)
-
-RoadMessage = _reflection.GeneratedProtocolMessageType('RoadMessage', (_message.Message,), {
-  'DESCRIPTOR' : _ROADMESSAGE,
-  '__module__' : 'events.events_pb2'
-  # @@protoc_insertion_point(class_scope:robocar.events.RoadMessage)
-  })
-_sym_db.RegisterMessage(RoadMessage)
-
-Point = _reflection.GeneratedProtocolMessageType('Point', (_message.Message,), {
-  'DESCRIPTOR' : _POINT,
-  '__module__' : 'events.events_pb2'
-  # @@protoc_insertion_point(class_scope:robocar.events.Point)
-  })
-_sym_db.RegisterMessage(Point)
-
-Ellipse = _reflection.GeneratedProtocolMessageType('Ellipse', (_message.Message,), {
-  'DESCRIPTOR' : _ELLIPSE,
-  '__module__' : 'events.events_pb2'
-  # @@protoc_insertion_point(class_scope:robocar.events.Ellipse)
-  })
-_sym_db.RegisterMessage(Ellipse)
-
-RecordMessage = _reflection.GeneratedProtocolMessageType('RecordMessage', (_message.Message,), {
-  'DESCRIPTOR' : _RECORDMESSAGE,
-  '__module__' : 'events.events_pb2'
-  # @@protoc_insertion_point(class_scope:robocar.events.RecordMessage)
-  })
-_sym_db.RegisterMessage(RecordMessage)
+DESCRIPTOR = _descriptor_pool.Default().AddSerializedFile(b'\n\x13\x65vents/events.proto\x12\x0erobocar.events\x1a\x1fgoogle/protobuf/timestamp.proto\"T\n\x08\x46rameRef\x12\x0c\n\x04name\x18\x01 \x01(\t\x12\n\n\x02id\x18\x02 \x01(\t\x12.\n\ncreated_at\x18\x03 \x01(\x0b\x32\x1a.google.protobuf.Timestamp\"C\n\x0c\x46rameMessage\x12$\n\x02id\x18\x01 \x01(\x0b\x32\x18.robocar.events.FrameRef\x12\r\n\x05\x66rame\x18\x02 \x01(\x0c\"d\n\x0fSteeringMessage\x12\x10\n\x08steering\x18\x01 \x01(\x02\x12\x12\n\nconfidence\x18\x02 \x01(\x02\x12+\n\tframe_ref\x18\x03 \x01(\x0b\x32\x18.robocar.events.FrameRef\"d\n\x0fThrottleMessage\x12\x10\n\x08throttle\x18\x01 \x01(\x02\x12\x12\n\nconfidence\x18\x02 \x01(\x02\x12+\n\tframe_ref\x18\x03 \x01(\x0b\x32\x18.robocar.events.FrameRef\"A\n\x10\x44riveModeMessage\x12-\n\ndrive_mode\x18\x01 \x01(\x0e\x32\x19.robocar.events.DriveMode\"f\n\x0eObjectsMessage\x12\'\n\x07objects\x18\x01 \x03(\x0b\x32\x16.robocar.events.Object\x12+\n\tframe_ref\x18\x02 \x01(\x0b\x32\x18.robocar.events.FrameRef\"\x80\x01\n\x06Object\x12(\n\x04type\x18\x01 \x01(\x0e\x32\x1a.robocar.events.TypeObject\x12\x0c\n\x04left\x18\x02 \x01(\x05\x12\x0b\n\x03top\x18\x03 \x01(\x05\x12\r\n\x05right\x18\x04 \x01(\x05\x12\x0e\n\x06\x62ottom\x18\x05 \x01(\x05\x12\x12\n\nconfidence\x18\x06 \x01(\x02\"&\n\x13SwitchRecordMessage\x12\x0f\n\x07\x65nabled\x18\x01 \x01(\x08\"\x8c\x01\n\x0bRoadMessage\x12&\n\x07\x63ontour\x18\x01 \x03(\x0b\x32\x15.robocar.events.Point\x12(\n\x07\x65llipse\x18\x02 \x01(\x0b\x32\x17.robocar.events.Ellipse\x12+\n\tframe_ref\x18\x03 \x01(\x0b\x32\x18.robocar.events.FrameRef\"\x1d\n\x05Point\x12\t\n\x01x\x18\x01 \x01(\x05\x12\t\n\x01y\x18\x02 \x01(\x05\"r\n\x07\x45llipse\x12%\n\x06\x63\x65nter\x18\x01 \x01(\x0b\x32\x15.robocar.events.Point\x12\r\n\x05width\x18\x02 \x01(\x05\x12\x0e\n\x06height\x18\x03 \x01(\x05\x12\r\n\x05\x61ngle\x18\x04 \x01(\x02\x12\x12\n\nconfidence\x18\x05 \x01(\x02\"\x82\x01\n\rRecordMessage\x12+\n\x05\x66rame\x18\x01 \x01(\x0b\x32\x1c.robocar.events.FrameMessage\x12\x31\n\x08steering\x18\x02 \x01(\x0b\x32\x1f.robocar.events.SteeringMessage\x12\x11\n\trecordSet\x18\x03 \x01(\t*-\n\tDriveMode\x12\x0b\n\x07INVALID\x10\x00\x12\x08\n\x04USER\x10\x01\x12\t\n\x05PILOT\x10\x02*2\n\nTypeObject\x12\x07\n\x03\x41NY\x10\x00\x12\x07\n\x03\x43\x41R\x10\x01\x12\x08\n\x04\x42UMP\x10\x02\x12\x08\n\x04PLOT\x10\x03\x42\nZ\x08./eventsb\x06proto3')
 
+_builder.BuildMessageAndEnumDescriptors(DESCRIPTOR, globals())
+_builder.BuildTopDescriptorsAndMessages(DESCRIPTOR, 'events.events_pb2', globals())
+if _descriptor._USE_C_DESCRIPTORS == False:
 
   DESCRIPTOR._options = None
+  DESCRIPTOR._serialized_options = b'Z\010./events'
+  _DRIVEMODE._serialized_start=1196
+  _DRIVEMODE._serialized_end=1241
+  _TYPEOBJECT._serialized_start=1243
+  _TYPEOBJECT._serialized_end=1293
+  _FRAMEREF._serialized_start=72
+  _FRAMEREF._serialized_end=156
+  _FRAMEMESSAGE._serialized_start=158
+  _FRAMEMESSAGE._serialized_end=225
+  _STEERINGMESSAGE._serialized_start=227
+  _STEERINGMESSAGE._serialized_end=327
+  _THROTTLEMESSAGE._serialized_start=329
+  _THROTTLEMESSAGE._serialized_end=429
+  _DRIVEMODEMESSAGE._serialized_start=431
+  _DRIVEMODEMESSAGE._serialized_end=496
+  _OBJECTSMESSAGE._serialized_start=498
+  _OBJECTSMESSAGE._serialized_end=600
+  _OBJECT._serialized_start=603
+  _OBJECT._serialized_end=731
+  _SWITCHRECORDMESSAGE._serialized_start=733
+  _SWITCHRECORDMESSAGE._serialized_end=771
+  _ROADMESSAGE._serialized_start=774
+  _ROADMESSAGE._serialized_end=914
+  _POINT._serialized_start=916
+  _POINT._serialized_end=945
+  _ELLIPSE._serialized_start=947
+  _ELLIPSE._serialized_end=1061
+  _RECORDMESSAGE._serialized_start=1064
+  _RECORDMESSAGE._serialized_end=1194
 # @@protoc_insertion_point(module_scope)

main.py

@@ -0,0 +1,274 @@
+#!/usr/bin/env python3
+
+from pathlib import Path
+
+import cv2
+import numpy as np
+import depthai as dai
+import east
+import blobconverter
+
+class HostSeqSync:
+    def __init__(self):
+        self.imfFrames = []
+    def add_msg(self, msg):
+        self.imfFrames.append(msg)
+    def get_msg(self, target_seq):
+        for i, imgFrame in enumerate(self.imfFrames):
+            if target_seq == imgFrame.getSequenceNum():
+                self.imfFrames = self.imfFrames[i:]
+                break
+        return self.imfFrames[0]
+
+pipeline = dai.Pipeline()
+version = "2021.2"
+pipeline.setOpenVINOVersion(version=dai.OpenVINO.Version.VERSION_2021_2)
+
+colorCam = pipeline.create(dai.node.ColorCamera)
+colorCam.setPreviewSize(256, 256)
+colorCam.setVideoSize(1024, 1024) # 4 times larger in both axis
+colorCam.setResolution(dai.ColorCameraProperties.SensorResolution.THE_1080_P)
+colorCam.setInterleaved(False)
+colorCam.setBoardSocket(dai.CameraBoardSocket.RGB)
+colorCam.setFps(10)
+
+controlIn = pipeline.create(dai.node.XLinkIn)
+controlIn.setStreamName('control')
+controlIn.out.link(colorCam.inputControl)
+
+cam_xout = pipeline.create(dai.node.XLinkOut)
+cam_xout.setStreamName('video')
+colorCam.video.link(cam_xout.input)
+
+# ---------------------------------------
+# 1st stage NN - text-detection
+# ---------------------------------------
+
+nn = pipeline.create(dai.node.NeuralNetwork)
+nn.setBlobPath(blobconverter.from_zoo(name="east_text_detection_256x256",zoo_type="depthai",shaves=6, version=version))
+colorCam.preview.link(nn.input)
+
+nn_xout = pipeline.create(dai.node.XLinkOut)
+nn_xout.setStreamName('detections')
+nn.out.link(nn_xout.input)
+
+# ---------------------------------------
+# 2nd stage NN - text-recognition-0012
+# ---------------------------------------
+
+manip = pipeline.create(dai.node.ImageManip)
+manip.setWaitForConfigInput(True)
+
+manip_img = pipeline.create(dai.node.XLinkIn)
+manip_img.setStreamName('manip_img')
+manip_img.out.link(manip.inputImage)
+
+manip_cfg = pipeline.create(dai.node.XLinkIn)
+manip_cfg.setStreamName('manip_cfg')
+manip_cfg.out.link(manip.inputConfig)
+
+manip_xout = pipeline.create(dai.node.XLinkOut)
+manip_xout.setStreamName('manip_out')
+
+nn2 = pipeline.create(dai.node.NeuralNetwork)
+nn2.setBlobPath(blobconverter.from_zoo(name="text-recognition-0012", shaves=6, version=version))
+nn2.setNumInferenceThreads(2)
+manip.out.link(nn2.input)
+manip.out.link(manip_xout.input)
+
+nn2_xout = pipeline.create(dai.node.XLinkOut)
+nn2_xout.setStreamName("recognitions")
+nn2.out.link(nn2_xout.input)
+
+def to_tensor_result(packet):
+    return {
+        name: np.array(packet.getLayerFp16(name))
+        for name in [tensor.name for tensor in packet.getRaw().tensors]
+    }
+
+def to_planar(frame):
+    return frame.transpose(2, 0, 1).flatten()
+
+with dai.Device(pipeline) as device:
+    q_vid = device.getOutputQueue("video", 4, blocking=False)
+    # This should be set to block, but would get to some extreme queuing/latency!
+    q_det = device.getOutputQueue("detections", 4, blocking=False)
+
+    q_rec = device.getOutputQueue("recognitions", 4, blocking=True)
+
+    q_manip_img = device.getInputQueue("manip_img")
+    q_manip_cfg = device.getInputQueue("manip_cfg")
+    q_manip_out = device.getOutputQueue("manip_out", 4, blocking=False)
+
+    controlQueue = device.getInputQueue('control')
+
+    frame = None
+    cropped_stacked = None
+    rotated_rectangles = []
+    rec_pushed = 0
+    rec_received = 0
+    host_sync = HostSeqSync()
+
+    class CTCCodec(object):
+        """ Convert between text-label and text-index """
+        def __init__(self, characters):
+            # characters (str): set of the possible characters.
+            dict_character = list(characters)
+
+            self.dict = {}
+            for i, char in enumerate(dict_character):
+                self.dict[char] = i + 1
+
+            self.characters = dict_character
+            #print(self.characters)
+            #input()
+        def decode(self, preds):
+            """ convert text-index into text-label. """
+            texts = []
+            index = 0
+            # Select max probabilty (greedy decoding) then decode index to character
+            preds = preds.astype(np.float16)
+            preds_index = np.argmax(preds, 2)
+            preds_index = preds_index.transpose(1, 0)
+            preds_index_reshape = preds_index.reshape(-1)
+            preds_sizes = np.array([preds_index.shape[1]] * preds_index.shape[0])
+
+            for l in preds_sizes:
+                t = preds_index_reshape[index:index + l]
+
+                # NOTE: t might be zero size
+                if t.shape[0] == 0:
+                    continue
+
+                char_list = []
+                for i in range(l):
+                    # removing repeated characters and blank.
+                    if not (i > 0 and t[i - 1] == t[i]):
+                        if self.characters[t[i]] != '#':
+                            char_list.append(self.characters[t[i]])
+                text = ''.join(char_list)
+                texts.append(text)
+
+                index += l
+
+            return texts
+
+    characters = '0123456789abcdefghijklmnopqrstuvwxyz#'
+    codec = CTCCodec(characters)
+
+    ctrl = dai.CameraControl()
+    ctrl.setAutoFocusMode(dai.CameraControl.AutoFocusMode.CONTINUOUS_VIDEO)
+    ctrl.setAutoFocusTrigger()
+    controlQueue.send(ctrl)
+
+    while True:
+        vid_in = q_vid.tryGet()
+        if vid_in is not None:
+            host_sync.add_msg(vid_in)
+
+        # Multiple recognition results may be available, read until queue is empty
+        while True:
+            in_rec = q_rec.tryGet()
+            if in_rec is None:
+                break
+            rec_data = bboxes = np.array(in_rec.getFirstLayerFp16()).reshape(30,1,37)
+            decoded_text = codec.decode(rec_data)[0]
+            pos = rotated_rectangles[rec_received]
+            print("{:2}: {:20}".format(rec_received, decoded_text),
+                "center({:3},{:3}) size({:3},{:3}) angle{:5.1f} deg".format(
+                    int(pos[0][0]), int(pos[0][1]), pos[1][0], pos[1][1], pos[2]))
+            # Draw the text on the right side of 'cropped_stacked' - placeholder
+            if cropped_stacked is not None:
+                cv2.putText(cropped_stacked, decoded_text,
+                                (120 + 10 , 32 * rec_received + 24),
+                                cv2.FONT_HERSHEY_SIMPLEX, 0.8, (0,255,0), 2)
+                cv2.imshow('cropped_stacked', cropped_stacked)
+            rec_received += 1
+
+        if cv2.waitKey(1) == ord('q'):
+            break
+
+        if rec_received >= rec_pushed:
+            in_det = q_det.tryGet()
+            if in_det is not None:
+                frame = host_sync.get_msg(in_det.getSequenceNum()).getCvFrame().copy()
+
+                scores, geom1, geom2 = to_tensor_result(in_det).values()
+                scores = np.reshape(scores, (1, 1, 64, 64))
+                geom1 = np.reshape(geom1, (1, 4, 64, 64))
+                geom2 = np.reshape(geom2, (1, 1, 64, 64))
+
+                bboxes, confs, angles = east.decode_predictions(scores, geom1, geom2)
+                boxes, angles = east.non_max_suppression(np.array(bboxes), probs=confs, angles=np.array(angles))
+                rotated_rectangles = [
+                    east.get_cv_rotated_rect(bbox, angle * -1)
+                    for (bbox, angle) in zip(boxes, angles)
+                ]
+
+                rec_received = 0
+                rec_pushed = len(rotated_rectangles)
+                if rec_pushed:
+                    print("====== Pushing for recognition, count:", rec_pushed)
+                cropped_stacked = None
+                for idx, rotated_rect in enumerate(rotated_rectangles):
+                    # Detections are done on 256x256 frames, we are sending back 1024x1024
+                    # That's why we multiply center and size values by 4
+                    rotated_rect[0][0] = rotated_rect[0][0] * 4
+                    rotated_rect[0][1] = rotated_rect[0][1] * 4
+                    rotated_rect[1][0] = rotated_rect[1][0] * 4
+                    rotated_rect[1][1] = rotated_rect[1][1] * 4
+
+                    # Draw detection crop area on input frame
+                    points = np.int0(cv2.boxPoints(rotated_rect))
+                    print(rotated_rect)
+                    cv2.polylines(frame, [points], isClosed=True, color=(255, 0, 0), thickness=1, lineType=cv2.LINE_8)
+
+                    # TODO make it work taking args like in OpenCV:
+                    # rr = ((256, 256), (128, 64), 30)
+                    rr = dai.RotatedRect()
+                    rr.center.x    = rotated_rect[0][0]
+                    rr.center.y    = rotated_rect[0][1]
+                    rr.size.width  = rotated_rect[1][0]
+                    rr.size.height = rotated_rect[1][1]
+                    rr.angle       = rotated_rect[2]
+                    cfg = dai.ImageManipConfig()
+                    cfg.setCropRotatedRect(rr, False)
+                    cfg.setResize(120, 32)
+                    # Send frame and config to device
+                    if idx == 0:
+                        w,h,c = frame.shape
+                        imgFrame = dai.ImgFrame()
+                        imgFrame.setData(to_planar(frame))
+                        imgFrame.setType(dai.ImgFrame.Type.BGR888p)
+                        imgFrame.setWidth(w)
+                        imgFrame.setHeight(h)
+                        q_manip_img.send(imgFrame)
+                    else:
+                        cfg.setReusePreviousImage(True)
+                    q_manip_cfg.send(cfg)
+
+                    # Get manipulated image from the device
+                    transformed = q_manip_out.get().getCvFrame()
+
+                    rec_placeholder_img = np.zeros((32, 200, 3), np.uint8)
+                    transformed = np.hstack((transformed, rec_placeholder_img))
+                    if cropped_stacked is None:
+                        cropped_stacked = transformed
+                    else:
+                        cropped_stacked = np.vstack((cropped_stacked, transformed))
+
+        if cropped_stacked is not None:
+            cv2.imshow('cropped_stacked', cropped_stacked)
+
+        if frame is not None:
+            cv2.imshow('frame', frame)
+
+        key = cv2.waitKey(1)
+        if  key == ord('q'):
+            break
+        elif key == ord('t'):
+            print("Autofocus trigger (and disable continuous)")
+            ctrl = dai.CameraControl()
+            ctrl.setAutoFocusMode(dai.CameraControl.AutoFocusMode.AUTO)
+            ctrl.setAutoFocusTrigger()
+            controlQueue.send(ctrl)

requirements.txt

@@ -5,3 +5,5 @@ opencv-python~=4.5.5.62
 google~=3.0.0
 google-api-core~=2.4.0
 setuptools==60.5.0
+protobuf3
+blobconverter>=1.2.9