The official implementation for the AAAI 2025 paper [Less Is More: Token Context-Aware Learning for Object Tracking].
conda create -n lmtrack python=3.8
conda activate lmtrack
bash install.sh
Put the tracking datasets in ./data. It should look like:
${PROJECT_ROOT}
-- data
-- lasot
|-- airplane
|-- basketball
|-- bear
...
-- got10k
|-- test
|-- train
|-- val
-- coco
|-- annotations
|-- images
-- trackingnet
|-- TRAIN_0
|-- TRAIN_1
...
|-- TRAIN_11
|-- TEST
Run the following command to set paths for this project
python tracking/create_default_local_file.py --workspace_dir . --data_dir ./data --save_dir ./output
After running this command, you can also modify paths by editing these two files
lib/train/admin/local.py # paths about training
lib/test/evaluation/local.py # paths about testing
Download pre-trained MAE ViT-Base weights and put it under $PROJECT_ROOT$/pretrained_models (different pretrained models can also be used, see MAE for more details).
python tracking/train.py \
--script lmtrack --config baseline \
--save_dir ./output \
--mode multiple --nproc_per_node 4 \
--use_wandb 1
Replace --config with the desired model config under experiments/lmtrack.
We use wandb to record detailed training logs, in case you don't want to use wandb, set --use_wandb 0.
- LaSOT or other off-line evaluated benchmarks (modify
--datasetcorrespondingly)
python tracking/test.py lmtrack baseline --dataset lasot --runid 300 --threads 8 --num_gpus 2
python tracking/analysis_results.py # need to modify tracker configs and names
- GOT10K-test
python tracking/test.py lmtrack baseline_384 --dataset got10k_test --runid 100 --threads 8 --num_gpus 2
python lib/test/utils/transform_got10k.py --tracker_name lmtrack --cfg_name baseline_384
- TrackingNet
python tracking/test.py lmtrack baseline_fullx4_384 --dataset trackingnet --runid 300 --threads 8 --num_gpus 2
python lib/test/utils/transform_trackingnet.py --tracker_name lmtrack --cfg_name baseline_fullx4_384