/AIC22_MTMC_ali

alibaba 2022 AICityChallenge No3. MTMCT. CityFlowV2 dataset.

Primary LanguagePython

The code is clone from AICITY2022_Track1_TAG

AICITY2022_Track1_TAG

The code is modified from AIC21-MTMC.

Install dependencies

cd ./detector/yolov5/ && pip install -r requirements.txt

Data & Models Preparation

  1. Download dataset from here, unzip it and put it in datasets folder. Thencd detector and run python gen_images_aic.py aic.yml to generate images.
  2. cd ./reid_training and follow the README_ReID.md to train the ReID models. Already-trained models can be downloaded from here. Put ReID models into the folder ./reid_bidir/reid_model/.
  3. Download yolov5x model (pretrained on COCO), and put it into the folder ./detector/yolov5/. Then cd ./detector/yolov5/, run:bash gen_det.sh aic.yml. Already-generated detection results are provided here.

MCMVT Results Generation

Run bash MCMVT.sh to generate MCMVT results. The result file is located in ./reid_bidir/reid-matching/tools/, the filename is defined in aic.yml.

You can modify params of yml files in ./config, ensure all paths are correct and can be found.