GStreamer: unable to start pipeline! about tensorrt_demos

[jkjung-avt]

Maybe GStreamer's v4l2src does not work for the USB webcam you're using. Please try if replace open_cam_usb() function with the following works.

def open_cam_usb(dev, width, height):
    return cv2.VideoCapture(dev)

[gzz1529657064]

Maybe GStreamer's v4l2src does not work for the USB webcam you're using. Please try if replace open_cam_usb() function with the following works.

def open_cam_usb(dev, width, height):
    return cv2.VideoCapture(dev)

thank you replay, you suggestion works!

[gzz1529657064]

I read your github repository 'tf_trt_models'. I want to transform the mobilenet_v1_yolo_v3 tensorflow model, but it need a mobilenet_v1_yolo_v3_egohands.config. I don't know how to write this file.
What should I do?

I feel that it is not easy to accelerate the YOLOv3 tensorflow model with tensorRT. I'm going to try caffe on Jetson TX2。Your blog has helped me a lot。Thank you very much!

[jkjung-avt]

I think you need to train your own 'mobilenet_v1_yolo_v3_egohands.config' model. You could refer to my Training a Hand Detector with TensorFlow Object Detection API blog post. I demonstrated how to train SSD and Faster RCNN models there.

Since this question is not directly related to 'tensorrt_demos', I'd appreciate it if you post further comments/questions in the more proper places, e.g. Disqus forum on my blog post or my GitHub hand-detection-tutorial repository.