Hi,
I've recently setup my NCS and am attempting to use the stream_infer program with a USB camera. I'm running Ubuntu 16.04 on a SurfaceBook.
When I launch the program via python3 stream_infer the video stream works, but the classification at the bottom of the display screen does not update from "toilet tissue, toilet paper, bathroom tissue (nan%)". Which happens to be the last classification listed in the categories.txt file.
I've searched through the stream_infer.py file, read the readme, and un-commented the:
print("loading %s : %s" % (preprocessed_image_buf.shape, preprocessed_image_buf ))
line in the input_thread section and found that an image is being preprocessed, as I get the preprocessed_image_buf.shape array as well as the preprocessed_image_buf matrix to print to the terminal.
I've also confirmed that the required file structures exist in the SqueezeNet directory, and the stream_infer program is able to reach that location and loads the stat.txt and categories.txt files correctly. I'm unable to determine where the inputsize.txt file is utilized though.
Any thoughts on where I should focus my troubleshooting efforts?
