This is the implementation of MTAtrack.
This code is based on DETR: [link].
d8519df08d38a186718a987f5974f017.mp4
- PyTorch 1.7.0
- torchvision 0.8.1
- cudatoolkit 10.2
- easydict
- cython
${MTAtrack_ROOT}
-- data
-- Lasot
|-- airplane
|-- basketball
|-- bear
...
-- GOT10k
|-- test
|-- train
|-- val
-- COCO
|-- annotations
|-- images
-- TrackingNet
|-- TRAIN_0
|-- TRAIN_1
...
|-- TRAIN_11
|-- TEST
- Download pre-trained weights into
./checpoints
[weights]
- paths setting
python tracking/create_default_local_file.py --workspace_dir . --data_dir ./data --save_dir .
- TrackingNet validation set
python tracking/test.py MTAtrack_st --dataset trackingnet --threads 16
python lib/test/utils/transform_trackingnet.py --tracker_name MTAtrack_st
- Lasot validation set
python tracking/test.py MTAtrack_st --dataset lasot --threads 16
python tracking/analysis_results.py # need to modify tracker configs and names
- GOT-10k validation set
python tracking/test.py MTAtrack_st --dataset got10k --threads 16
python lib/test/utils/transform_trackingnet.py --tracker_name MTAtrack_st
- UAV123 validation set
python tracking/test.py MTAtrack_st baseline --dataset uav --threads 16
python tracking/analysis_results.py # need to modify tracker configs and names
- OTB100 validation set
python tracking/test.py MTAtrack_st baseline --dataset otb --threads 16
python tracking/analysis_results.py # need to modify tracker configs and names
-
TrackingNet results
[trackingnet] -
Lasot results
[Lasot] -
GOT-10k results
[GOT-10k] -
UAV123 results
[UAV123] -
OTB100 results
[OTB100]