Demo

Here's a GIF demonstrating the detection results when running at 35 fps on a Jetson Nano (GIF has reduced fps due to file size limitations).

Dependencies

python3 train.py -tr <path to training .csv> -te <path to testing .csv> -w <name of saved weightfile>

Images should be listed in .csv files for training and testing, respectively. Each image should be given as a line:

<path to image>, <class>

python3 video_detection.py -p <path to video> -w <name of weightfile> -v <visual output (0 or 1)>

A sample clip is provided for demo purposes.

python3 speed_benchmark.py -p <path to video> -w <name of weightfile>