README.md

Environment Installation

conda create -n sts python=3.8
conda activate sts
bash install.sh

Project Paths Setup

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

Data Preparation

Put the tracking datasets in ./data. It should look like:

${PROJECT_ROOT}
  -- data
      -- lasher
          |-- trainingset
          |-- testingset
          |-- trainingsetList.txt
          |-- testingsetList.txt
          ...

Download LasHeR,VTUAV,RGBT234.

Training

Training from scrath

Download OSTrack. And, training with dataset Lasot, COCO, GOT-10k and TrackingNet for SOT pretrained model. （Download SOT pretrained weights） And put it under $PROJECT_ROOT$/pretrained_models.

python tracking/train.py --script select_track --config vitb_256_select_32x1_1e4_lasher_15ep_sot --save_dir ./output --mode multiple --nproc_per_node 4

Evaluation

Download checkpoint and put it under $PROJECT_ROOT$/output.

python tracking/test.py select_track vitb_256_select_32x1_1e4_lasher_15ep_sot --dataset_name lasher_test

Download raw result and put it under $PROJECT_ROOT$/output.

python tracking/analysis_results.py

We refer you to LasHeR Toolkit for LasHeR, RGBT234 and RGBT210 evaluation, and refer you to MPR_MSR_Evaluation for VTUAV evaluation.

Result

Dataset	Model	Backbone	Pretraining	Precision	NormPrec	Success	FPS
LasHeR	MRTTrack	ViT-Base	SOT	70.2	66.5	56.5	32.5

Dataset	Model	Backbone	Pretraining	Precision	Success	FPS
RGBT210	MRTTrack	ViT-Base	SOT	85.6	63.1	32.5
RGBT234	MRTTrack	ViT-Base	SOT	87.2	64.1	32.5

Dataset	Model	Backbone	Pretraining	MPR	MSR	FPS
VTUAV-ST	MRTTrack	ViT-Base	SOT	84.3	72.1	32.5

Acknowledgments

Our project is developed upon TBSI. Thanks for their contributions which help us to quickly implement our ideas.