SiamDMU
This is the official implementation with training code for 'SiamDMU: Siamese Dual Mask Update Networkfor Visual Object Tracking'.
Introduction
We propose a novel tracker named Siamese Dual Mask Update (SiamDMU), which utilizes motion and semantic information to generate the enhanced tracking results for updating the template.
The Results of SiamDMU are here. Baidu:2333
Requirements
- Ubuntu 20.04
- Pytorch 1.3.1
- Python 3.7
Installation
Please refer to PySOT_INSTALL.md, FlowNet_README.md and DeepMask_README.md for installation.
You can also download the code of SiamDMU from BaiduYun password: ofed. This file has included the model of SiamRPN++, FlowNet and DeepMask in the dirctory.
Usage
Download models
- Please download the SiamRPN++ and SiamRPN++_otb model.pth to the path './experiments/siamrpn_r50_l234_dwxcorr/' and './experiments/siamrpn_r50_l234_dwxcorr_otb/', respectively.
- Download the FlowNet2-C_checkpoint.pth.tar to the './FlowWrapping/pretrained_model/'.
- Download the DeepMask.pth.tar to the './deepmask/pretrained/'.
Test
- Modify the dataset path 'dataset_root'.
- run the './tools/test_SiamDMU_VOT.py'.
Train
- run the './updatenet/create_template.py'.
- run the './updatenet/train_upd.py'.