Inference images using Hailo

General
1

I’m using RPI with Hailo AI kit (version 4.20.0)

I’m trying to inference an image and getting the following error:

[HailoRT] [error] CHECK failed - Memory size of vstream yolov8n/input_layer1 does not match the frame count! (Expected 25165824000, got 19660800)
[HailoRT] [error] CHECK_SUCCESS failed with status=HAILO_INVALID_ARGUMENT(2)

this is my code:

 imageDetections = []
        file = request.files['image']
        image = Image.open(file)
        image_height = image.height
        image_width = image.width


        devices = Device.scan()
        with VDevice(device_ids=devices) as target:
            hef = HEF(f'resources/{model}.hef')
            utils = ObjectDetectionUtils(f"resources/{model}.txt")


            # Convert the image to a numpy array
            image = image.resize((1280, 1280))
            image = np.array(image)

            network_group = configure_and_get_network_group(hef, target)
            network_group_params = network_group.create_params()
            
            input_vstreams_params, output_vstreams_params = create_input_output_vstream_params(network_group)

            # print info of input & output
            input_vstream_info, output_vstream_info = print_input_output_vstream_info(hef)
            start_time = time.time()
            with InferVStreams(network_group, input_vstreams_params, output_vstreams_params) as infer_pipeline:
                total_inference_time = 0
                with network_group.activate(network_group_params):
                    np_array = np.array(image, dtype=np.float32) / 255.0
                    input_data = {input_vstream_info[0].name: np_array}

                    raw_image = infer_pipeline.infer(input_data)
                    detections = utils.extract_detections(raw_image[next(iter(output_vstream_info)).name][0], range)
                    #print(detections)
                    for detection in detections:
                        imageDetections.append(detection)
                    end_time = time.time()
                    total_inference_time = (end_time - start_time)
                    logger.info("Total inference time: {} sec", total_inference_time)

        #im = Image.open(BytesIO(base64.b64decode(data['image'])))
        return jsonify(imageDetections)

this is my hef file:

Architecture HEF was compiled for: HAILO8
Network group name: yolov8n, Single Context
    Network name: yolov8n/yolov8n
        VStream infos:
            Input  yolov8n/input_layer1 UINT8, NHWC(1280x1280x3)
            Output yolov8n/yolov8_nms_postprocess FLOAT32, HAILO NMS BY CLASS(number of classes: 2, maximum bounding boxes per class: 100, maximum frame size: 4008)
            Operation:
                Op YOLOV8
                Name: YOLOV8-Post-Process
                Score threshold: 0.200
                IoU threshold: 0.70
                Classes: 2
                Cross classes: false
                NMS results order: BY_CLASS
                Max bboxes per class: 100
                Image height: 1280
                Image width: 1280

Any idea what can be the problem? I tried to look for previous messages but i’m not ablt to find useful information

2

BTW → Using the same hef file with Gstreamer is working fine.

3

Found a similar concern here with assistance