This work was done when Chengxi Li was an intern at Honda Research Institute, San Jose, CA
The code is developed with CUDA 10.2, Python 3.6, PyTorch = 1.0.0
Step 1: Use Detectron.pytorch to generate detection.txt for each clip Step 2: Use deep_sort to generate detection features Step 3: Use deep_sort to generate tracking.txt Step 4: Use script 'refine_tracking' to generate refined_tracking.txt