rnudb / vot Goto Github PK

Author: Arnaud Baradat - SCIA 2024

Make sure to have installed the virtual environment with the requirements.txt file and have activated it.

To download the dataset, please get the MOT15 dataset from here. Then, put the MOT15 dataset in the data folder. (under the name "MOT15")

Please also add a outputs folder in the data folder. This is where the detections with associated scores will be stored. Finally, add a yolo_dets folder in the data folder. This is where the detections from YOLO will be stored. It should be in the same format as the MOT15 dataset. (only the train folder of MOT15 is needed as it is the only one used)

The folder structure should look like this:

data
└── MOT15
    ├── test
    │   ├── ADL-Rundle-6
    │   ├── ETH-Bahnhof
    │   ├── ETH-Pedcross2
    │   ├── KITTI-13
    │   ├── KITTI-17
    │   ├── PETS09-S2L1
    │   ├── TUD-Campus
    │   ├── TUD-Stadtmitte
    │   ├── Venice-2
    │   └── Venice-6
    └── train
        ├── ADL-Rundle-1
        ├── ADL-Rundle-3
        ├── AVSS-AB
        ├── AVSS-AV
        ├── ETH-Bahnhof
        ├── ETH-Pedcross2
        ├── ETH-Sunnyday
        ├── KITTI-13
        ├── KITTI-17
        ├── PETS09-S2L1
        ├── TUD-Campus
        ├── TUD-Stadtmitte
        ├── Venice-2
└── yolo_dets
    ├── train
        ├── ADL-Rundle-1
        ├── ADL-Rundle-3
        ├── AVSS-AB
        ├── AVSS-AV
        ├── ETH-Bahnhof
        ├── ETH-Pedcross2
        ├── ETH-Sunnyday
        ├── KITTI-13
        ├── KITTI-17
        ├── PETS09-S2L1
        ├── TUD-Campus
        ├── TUD-Stadtmitte
        ├── Venice-2

└── outputs

To run the code, please run the following command:

python main.py

This will run the code on the whole MOT15 training examples. To run the code on a different sequence, please change the path read in the file.

To re-run the detections with YOLO, please run the following command:

python yolo_detection.py

This will run the YOLO detections on the MOT15 training examples. To run the code on a different sequence, please change the path read in the file.

• Clone the MOTChallengeEvalKit
• Copy the data/outputs folder content to the data/trackers/mot-challenge/MOT15-train folder in the MOTChallengeEvalKit under the (new) directory name CustomTracker
• Run the following command:

python scripts/run_mot_challenge.py --BENCHMARK MOT15 --SPLIT_TO_EVAL train --TRACKERS_TO_EVAL CustomTracker --METRICS HOTA CLEAR Identity VACE --USE_PARALLEL False --NUM_PARALLEL_CORES 1 --DO_PREPROC False

Benmarks results are available in the benchmark folder.