Data
- static/ folder tree:
.
├── benchmarks
├── data_template
│ ├── gt
│ │ └── mot_challenge
│ │ ├── MOT16
│ │ └── seqmaps
│ └── trackers
│ └── mot_challenge
│ └── MOT16
│ └── ch_yolov5m_deep_sort
│ └── data
└── test_data
├── gt
└── pred
- Prepare:
- Put your all your gt files in test_data/gt folder, in folder utils, run:
python get_gt_file --root_path {path contain all your data} --save_path {static/test_data/gt}- Put your prediction files in test_data/pred folder
- Rename your prediction files the same as gt folder. eg: gt/NVR-CH01... -> pred/NVR-CH01
- Notes:
- Name of prediction and groundtruth files must be matched
Evaluate:
- In create_data.py, modify line 47 to path of folder contain all videos
- In mot_eval/ folder, run:
bash create_mot_eval.sh {name of your save folder result eg. fairmot_dla34_finetune_dataset...}