Hi jakaskerl

How are you doing bro? I have a question regarding that project, I adjusted the project as I want and it's working as expected, now I have to test detection and tracking with crowd, currently I can't put camera to street or any public places, so I wanted to test it with videos. However, I couldn't manage to integrate the video reading in the project, I tried using this method https://docs.luxonis.com/projects/api/en/latest/samples/ObjectTracker/object_tracker_video/#source-code
In the beginnign I tried to add this to my code but I couldn't get it right, so I added Yolo to the code in the documentation I mentioned above. But it's not running on device, how can I test my project with video? can you help me with that please.

Thanks,
Fakhrullo

Hi jakaskerl

I added manip mode as you said, I added like this:
# Creating Manip node
manip = pipeline.create(dai.node.ImageManip)
# Setting CropRect for the Region of Interest
manip.initialConfig.setCropRect(*custom_roi)
# Setting Resize for the neural network input size
manip.initialConfig.setResize(640, 640)
# Setting maximum output frame size based on the desired output dimensions
max_output_width = 640
max_output_height = 640
max_output_frame_size = 3 * max_output_width * max_output_height # Assuming 3 channels for BGR image
manip.setMaxOutputFrameSize(max_output_frame_size)
# Connecting Manip node to ColorCamera
camRgb.preview.link(manip.inputImage)
# Connecting Manip node to YoloDetectionNetwork
manip.out.link(detectionNetwork.input)

But I'm getting black screen, Where am I making mistake? is there any tutorial or documentation how to do that?

Hi @Fakhrullo
Can I see the full code?
If you are inputting the preview image into the manip, make sure it's larger than the output image, otherwise the crop makes no sense.
Maybe you could link the .video output and set the frame type to RGB.

Thanks,
Jaka

    Hi jakaskerl
    Here's the code

    from pathlib import Path

    import cv2

    import depthai as dai

    import time

    from environs import Env
    env = Env()env.read_env()
    MxID = env('MxID')# Set your custom ROI coordinates (x, y, width, height)custom_roi = (350, 250, 640, 640) # Example coordinates, adjust as needed
    # tiny yolo v4 label textslabelMap = [ "person", "bicycle", "car", "motorbike", "aeroplane", "bus", "train", "truck", "boat", "traffic light", "fire hydrant", "stop sign", "parking meter", "bench", "bird", "cat", "dog", "horse", "sheep", "cow", "elephant", "bear", "zebra", "giraffe", "backpack", "umbrella", "handbag", "tie", "suitcase", "frisbee", "skis", "snowboard", "sports ball", "kite", "baseball bat", "baseball glove", "skateboard", "surfboard", "tennis racket", "bottle", "wine glass", "cup", "fork", "knife", "spoon", "bowl", "banana", "apple", "sandwich", "orange", "broccoli", "carrot", "hot dog", "pizza", "donut", "cake", "chair", "sofa", "pottedplant", "bed", "diningtable", "toilet", "tvmonitor", "laptop", "mouse", "remote", "keyboard", "cell phone", "microwave", "oven", "toaster", "sink", "refrigerator", "book", "clock", "vase", "scissors", "teddy bear", "hair drier", "toothbrush"]
    nnPath = str((Path(__file__).parent / Path('model/yolov6n_coco_640x640_openvino_2022.1_6shave.blob')).resolve().absolute())
    # Create pipelinepipeline = dai.Pipeline()
    # Define sources and outputscamRgb = pipeline.create(dai.node.ColorCamera)detectionNetwork = pipeline.create(dai.node.YoloDetectionNetwork)objectTracker = pipeline.create(dai.node.ObjectTracker)
    xlinkOut = pipeline.create(dai.node.XLinkOut)trackerOut = pipeline.create(dai.node.XLinkOut)
    xlinkOut.setStreamName("preview")trackerOut.setStreamName("tracklets")
    # Creating Manip nodemanip = pipeline.create(dai.node.ImageManip)# Setting CropRect for the Region of Interestmanip.initialConfig.setCropRect(*custom_roi)# Setting Resize for the neural network input sizemanip.initialConfig.setResize(640, 640)# Setting maximum output frame size based on the desired output dimensionsmax_output_width = 640max_output_height = 640max_output_frame_size = 3 * max_output_width * max_output_height # Assuming 3 channels for BGR imagemanip.setMaxOutputFrameSize(max_output_frame_size)
    # Propertiesif MxID == "14442C10C1AD3FD700": camRgb.setImageOrientation(dai.CameraImageOrientation.HORIZONTAL_MIRROR)camRgb.setPreviewSize(640, 640)camRgb.setResolution(dai.ColorCameraProperties.SensorResolution.THE_1080_P)camRgb.setInterleaved(False)camRgb.setColorOrder(dai.ColorCameraProperties.ColorOrder.BGR)camRgb.setFps(40)
    # Network specific settingsdetectionNetwork.setConfidenceThreshold(0.5)detectionNetwork.setNumClasses(80)detectionNetwork.setCoordinateSize(4)# detectionNetwork.setAnchors([10, 14, 23, 27, 37, 58, 81, 82, 135, 169, 344, 319]) #YOLOv4 uchun# detectionNetwork.setAnchorMasks({"side26": [1, 2, 3], "side13": [3, 4, 5]})detectionNetwork.setAnchors([10,13, 16,30, 33,23, 30,61, 62,45, 59,119, 116,90, 156,198, 373,326]) #YOLOv5 uchundetectionNetwork.setAnchorMasks({"side52": [0,1,2], "side26": [3,4,5], "side13": [6,7,8]})detectionNetwork.setIouThreshold(0.5)detectionNetwork.setBlobPath(nnPath)detectionNetwork.setNumInferenceThreads(2)detectionNetwork.input.setBlocking(False)
    objectTracker.setDetectionLabelsToTrack([0]) # track only person# possible tracking types: ZERO_TERM_COLOR_HISTOGRAM, ZERO_TERM_IMAGELESS, SHORT_TERM_IMAGELESS, SHORT_TERM_KCFobjectTracker.setTrackerType(dai.TrackerType.ZERO_TERM_COLOR_HISTOGRAM)# take the smallest ID when new object is tracked, possible options: SMALLEST_ID, UNIQUE_IDobjectTracker.setTrackerIdAssignmentPolicy(dai.TrackerIdAssignmentPolicy.SMALLEST_ID)
    #Linking# Connecting Manip node to ColorCameracamRgb.preview.link(manip.inputImage)# Connecting Manip node to YoloDetectionNetworkmanip.out.link(detectionNetwork.input)# camRgb.preview.link(detectionNetwork.input)objectTracker.passthroughTrackerFrame.link(xlinkOut.input)

    detectionNetwork.passthrough.link(objectTracker.inputTrackerFrame)
    detectionNetwork.passthrough.link(objectTracker.inputDetectionFrame)detectionNetwork.out.link(objectTracker.inputDetections)objectTracker.out.link(trackerOut.input)
    device = dai.DeviceInfo(MxID)
    # Connect to device and start pipelinewith dai.Device(pipeline, device) as device:
    preview = device.getOutputQueue("preview", 4, False) tracklets = device.getOutputQueue("tracklets", 4, False)
    startTime = time.monotonic() counter = 0 fps = 0 frame = None
    while(True): imgFrame = preview.get() track = tracklets.get()
    counter+=1 current_time = time.monotonic() if (current_time - startTime) > 1 : fps = counter / (current_time - startTime) counter = 0 startTime = current_time
    color = (255, 0, 0) text_color = (0, 0, 255) rectangle = (111, 147, 26)
    frame = imgFrame.getCvFrame() trackletsData = track.tracklets for t in trackletsData: if t.status.name == "TRACKED": roi = t.roi.denormalize(frame.shape[1], frame.shape[0]) x1 = int(roi.topLeft().x) y1 = int(roi.topLeft().y) x2 = int(roi.bottomRight().x) y2 = int(roi.bottomRight().y)
    try: label = labelMap[t.label] except: label = t.label # if t.status.name == 'TRACKED': cv2.putText(frame, str(label), (x1 + 10, y1 + 20), cv2.FONT_HERSHEY_TRIPLEX, 0.5, text_color) cv2.putText(frame, f"ID: {[t.id]}", (x1 + 10, y1 + 45), cv2.FONT_HERSHEY_TRIPLEX, 0.5, text_color) cv2.putText(frame, t.status.name, (x1 + 10, y1 + 70), cv2.FONT_HERSHEY_TRIPLEX, 0.5, text_color) cv2.rectangle(frame, (x1, y1), (x2, y2), rectangle, cv2.FONT_HERSHEY_SIMPLEX)
    cv2.putText(frame, "FPS: {:.2f}".format(fps), (2, frame.shape[0] - 4), cv2.FONT_HERSHEY_TRIPLEX, 0.6, text_color)
    cv2.imshow("tracker", frame)
    if cv2.waitKey(1) == ord('q'): break

    Here's better version:

    from pathlib import Path
    import cv2
    import depthai as dai
    import time
    
    from environs import Env
    
    env = Env()
    env.read_env()
    
    MxID = env('MxID')
    # Set your custom ROI coordinates (x, y, width, height)
    custom_roi = (350, 250, 640, 640)  # Example coordinates, adjust as needed
    
    # tiny yolo v4 label texts
    labelMap = [
        "person",         "bicycle",    "car",           "motorbike",     "aeroplane",   "bus",           "train",
        "truck",          "boat",       "traffic light", "fire hydrant",  "stop sign",   "parking meter", "bench",
        "bird",           "cat",        "dog",           "horse",         "sheep",       "cow",           "elephant",
        "bear",           "zebra",      "giraffe",       "backpack",      "umbrella",    "handbag",       "tie",
        "suitcase",       "frisbee",    "skis",          "snowboard",     "sports ball", "kite",          "baseball bat",
        "baseball glove", "skateboard", "surfboard",     "tennis racket", "bottle",      "wine glass",    "cup",
        "fork",           "knife",      "spoon",         "bowl",          "banana",      "apple",         "sandwich",
        "orange",         "broccoli",   "carrot",        "hot dog",       "pizza",       "donut",         "cake",
        "chair",          "sofa",       "pottedplant",   "bed",           "diningtable", "toilet",        "tvmonitor",
        "laptop",         "mouse",      "remote",        "keyboard",      "cell phone",  "microwave",     "oven",
        "toaster",        "sink",       "refrigerator",  "book",          "clock",       "vase",          "scissors",
        "teddy bear",     "hair drier", "toothbrush"
    ]
    
    nnPath = str((Path(__file__).parent / Path('model/yolov6n_coco_640x640_openvino_2022.1_6shave.blob')).resolve().absolute())
    
    # Create pipeline
    pipeline = dai.Pipeline()
    
    # Define sources and outputs
    camRgb = pipeline.create(dai.node.ColorCamera)
    detectionNetwork = pipeline.create(dai.node.YoloDetectionNetwork)
    objectTracker = pipeline.create(dai.node.ObjectTracker)
    
    xlinkOut = pipeline.create(dai.node.XLinkOut)
    trackerOut = pipeline.create(dai.node.XLinkOut)
    
    xlinkOut.setStreamName("preview")
    trackerOut.setStreamName("tracklets")
    
    # Creating Manip node
    manip = pipeline.create(dai.node.ImageManip)
    # Setting CropRect for the Region of Interest
    manip.initialConfig.setCropRect(*custom_roi)
    # Setting Resize for the neural network input size
    manip.initialConfig.setResize(640, 640)
    # Setting maximum output frame size based on the desired output dimensions
    max_output_width = 640
    max_output_height = 640
    max_output_frame_size = 3 * max_output_width * max_output_height # Assuming 3 channels for BGR image
    manip.setMaxOutputFrameSize(max_output_frame_size)
    
    # Properties
    if MxID == "14442C10C1AD3FD700":
        camRgb.setImageOrientation(dai.CameraImageOrientation.HORIZONTAL_MIRROR)
    camRgb.setPreviewSize(640, 640)
    camRgb.setResolution(dai.ColorCameraProperties.SensorResolution.THE_1080_P)
    camRgb.setInterleaved(False)
    camRgb.setColorOrder(dai.ColorCameraProperties.ColorOrder.BGR)
    camRgb.setFps(40)
    
    # Network specific settings
    detectionNetwork.setConfidenceThreshold(0.5)
    detectionNetwork.setNumClasses(80)
    detectionNetwork.setCoordinateSize(4)
    # detectionNetwork.setAnchors([10, 14, 23, 27, 37, 58, 81, 82, 135, 169, 344, 319]) #YOLOv4 uchun
    # detectionNetwork.setAnchorMasks({"side26": [1, 2, 3], "side13": [3, 4, 5]})
    detectionNetwork.setAnchors([10,13, 16,30, 33,23, 30,61, 62,45, 59,119, 116,90, 156,198, 373,326]) #YOLOv5 uchun
    detectionNetwork.setAnchorMasks({"side52": [0,1,2], "side26": [3,4,5], "side13": [6,7,8]})
    detectionNetwork.setIouThreshold(0.5)
    detectionNetwork.setBlobPath(nnPath)
    detectionNetwork.setNumInferenceThreads(2)
    detectionNetwork.input.setBlocking(False)
    
    objectTracker.setDetectionLabelsToTrack([0])  # track only person
    # possible tracking types: ZERO_TERM_COLOR_HISTOGRAM, ZERO_TERM_IMAGELESS, SHORT_TERM_IMAGELESS, SHORT_TERM_KCF
    objectTracker.setTrackerType(dai.TrackerType.ZERO_TERM_COLOR_HISTOGRAM)
    # take the smallest ID when new object is tracked, possible options: SMALLEST_ID, UNIQUE_ID
    objectTracker.setTrackerIdAssignmentPolicy(dai.TrackerIdAssignmentPolicy.SMALLEST_ID)
    
    #Linking
    # Connecting Manip node to ColorCamera
    camRgb.preview.link(manip.inputImage)
    # Connecting Manip node to YoloDetectionNetwork
    manip.out.link(detectionNetwork.input)
    # camRgb.preview.link(detectionNetwork.input)
    objectTracker.passthroughTrackerFrame.link(xlinkOut.input)
    
    
    detectionNetwork.passthrough.link(objectTracker.inputTrackerFrame)
    
    detectionNetwork.passthrough.link(objectTracker.inputDetectionFrame)
    detectionNetwork.out.link(objectTracker.inputDetections)
    objectTracker.out.link(trackerOut.input)
    
    device = dai.DeviceInfo(MxID)
    
    # Connect to device and start pipeline
    with dai.Device(pipeline, device) as device:
    
        preview = device.getOutputQueue("preview", 4, False)
        tracklets = device.getOutputQueue("tracklets", 4, False)
    
        startTime = time.monotonic()
        counter = 0
        fps = 0
        frame = None
    
        while(True):
            imgFrame = preview.get()
            track = tracklets.get()
    
            counter+=1
            current_time = time.monotonic()
            if (current_time - startTime) > 1 :
                fps = counter / (current_time - startTime)
                counter = 0
                startTime = current_time
    
            color = (255, 0, 0)
            text_color = (0, 0, 255)
            rectangle = (111, 147, 26)
    
            frame = imgFrame.getCvFrame()
            trackletsData = track.tracklets
            for t in trackletsData:
                if t.status.name == "TRACKED":
                    roi = t.roi.denormalize(frame.shape[1], frame.shape[0])
                    x1 = int(roi.topLeft().x)
                    y1 = int(roi.topLeft().y)
                    x2 = int(roi.bottomRight().x)
                    y2 = int(roi.bottomRight().y)
    
                    try:
                        label = labelMap[t.label]
                    except:
                        label = t.label
                    # if t.status.name == 'TRACKED':
                    cv2.putText(frame, str(label), (x1 + 10, y1 + 20), cv2.FONT_HERSHEY_TRIPLEX, 0.5, text_color)
                    cv2.putText(frame, f"ID: {[t.id]}", (x1 + 10, y1 + 45), cv2.FONT_HERSHEY_TRIPLEX, 0.5, text_color)
                    cv2.putText(frame, t.status.name, (x1 + 10, y1 + 70), cv2.FONT_HERSHEY_TRIPLEX, 0.5, text_color)
                    cv2.rectangle(frame, (x1, y1), (x2, y2), rectangle, cv2.FONT_HERSHEY_SIMPLEX)
    
            cv2.putText(frame, "FPS: {:.2f}".format(fps), (2, frame.shape[0] - 4), cv2.FONT_HERSHEY_TRIPLEX, 0.6, text_color)
    
            cv2.imshow("tracker", frame)
    
            if cv2.waitKey(1) == ord('q'):
                break

      Hi Fakhrullo
      setCropRect() expects normalized values. Divide it by 416, then I believe it should work.

      Thanks,
      Jaka

        5 days later

        Hi jakaskerl
        Thanks it worked fine, I'm getting cropped frame.
        Is it possible to fuse cropped frame and whole frame? I mean it shows whole frame and cropped frame within whole frame may be with a box? If It is impossible how can I show whole and cropped frames in 2 different windows?

          Hi Fakhrullo
          Sorry, I don't really understand what you are trying to achieve. Both frames are numpy arrays which means you can easily stack one on top of the other.

          Thanks,
          Jaka

            Hi jakaskerl
            Yeah, basically I want to retrieve both full and cropped frames, to achieve that what should I do?

            Fakhrullo camRgb.preview.link(manip.inputImage)

            Here, you can pipe the preview stream to a new XLink node, so you can view it on host side.
            This will enable you to see both the full preview frame as well as cropped frame (on which the inference was made).

            Thanks,
            Jaka