FairMOT is a joint initiative of Huazhong University of Science and Technology and Microsoft Research Asia，Thesis AddressFairMOT: On the Fairness of Detection and Re-Identification in Multiple Object Tracking，This project is a further improvement on its basis to provide a theoretical basis for the research.
The original FairMOT model code is derived fromPaddleDetection，End-to-end vision suite based on Flying PaddlePaddle, covering target detection, instance segmentation, tracking, key point detection and other directions, providing a code base for model model research。
MOT17 public data is derived frommotchallenge，An important benchmark in the field of multi-objective tracking is MOTChallenge, which is a public platform for uploading and publishing research results of multi-objective tracking methods and has the largest publicly available pedestrian tracking dataset.
AI StudioIt is an AI learning and practical training community based on Baidu's deep learning open source platform Flying Paddle, which provides developers with a powerful online training environment, free GPU computing power and storage resources.LuGuThis is my community account, welcome to discuss and learn from the big guys。
- requirements.txt：Code to run the installation configuration file
- configs：Configuration files for model runs
- dataset：Dataset storage file
- ppdet：Model storage file containing the individual model blocks
- tools.：Training, Prediction files
- output：Weighted files, weighted downloads atGoogle Cloud Drive
- Installing dependencies
pip install -r requirements.txt
- Model Training
CUDA_VISIBLE_DEVICES=0
!python -u tools/train.py \
-c configs/fairmot/fairmot_dla34_30e_576x320.yml \
- Model Evaluation
CUDA_VISIBLE_DEVICES=0
!python tools/eval_mot.py \
-c configs/fairmot/fairmot_dla34_30e_1088x608.yml \
-o weights=output/T_724fairmot_dla34_30e_1088x608/model_final.pdparams