Skip to content

Latest commit

 

History

History
 
 

NanoTrack

README.md

NanoTrack

network

  • NanoTrack is a FCOS-style one-stage anchor-free object trakcing model which mainly referring to SiamBAN and LightTrack

  • NanoTrack is a simple, lightweight and high speed tracking network, and it is very suitable for deployment on embedded or mobile devices. We provide Android demo and MacOS demo based on ncnn inference framework. macs

  • We provide PyTorch code. It is very friendly for training with much lower GPU memory cost than other models. We only use GOT-10k as tranining set, and it only takes two hours to train on GPU3090

    NanoTrack  VOT2018 EAO 0.301

LightTrack VOT2018 EAO 0.42x

NanoTrack for PyTorch

  • Build
python  setup.py build_ext --inplace
  • Prepare data
1. cd xxx/xxx/NanoTrack 

2. mkdir data 
Download GOT-10k https://pan.baidu.com/s/10crE2uKR182fA93XRB3jyg password: 5ebm
How to crop GOT-10k https://pan.baidu.com/s/1ouqVMVAsLtXDWeanPYTkHw password: owlo
Put your training data into data directory 

3. mkdir datasets
Download VOT2018 https://pan.baidu.com/s/1MOWZ5lcxfF0wsgSuj5g4Yw password: e5eh  
Put your testing data into datasets directory
  • Train
python ./bin/train.py
  • Test
python ./bin/test.py
  • Eval
python ./bin/eval.py

checkpoint_e26.pth VOT2018  
------------------------------------------------------------
|Tracker Name| Accuracy | Robustness | Lost Number |  EAO  |
------------------------------------------------------------
| nanotrack  |  0.550   |   0.356    |    76.0     | 0.301 |
------------------------------------------------------------
  • Search params
python ./bin/hp_search.py
  • Calculate flops
python cal_macs_params.py
  • Calculate speed
python cal_speed.py
  • PyTorch to ONNX
python ./pytorch2onnx.py
  • ONNX to NCNN
https://convertmodel.com/

NanoTrack for MacOS

PC demo

  • Modify your own CMakeList.txt

  • Build (Apple M1 CPU)

    $ sh make_macos_arm64.sh

NanoTrack for Android

Android demo

  • Modify your own CMakeList.txt

  • Download(password: 6cdd) OpenCV and NCNN libraries for Android

Reference

https://github.com/Tencent/ncnn

https://github.com/hqucv/siamban

https://github.com/researchmm/LightTrack

https://github.com/Z-Xiong/LightTrack-ncnn

https://github.com/FeiGeChuanShu/ncnn_Android_LightTrack