siamdw.md

SiamDW tutorial

Testing

We assume the root path is $TracKit, e.g. /home/zpzhang/TracKit

Set up environment

Please follow readme of Ocean to install the environment.

Prepare data and models

1. Download the pretrained PyTorch model to $TracKit/snapshot.
2. Download json files of testing data and put thme in $TracKit/dataset.
3. Download testing data e.g. VOT2017 and put them in $TracKit/dataset.

Testing

In root path $TracKit,

python tracking/test_siamdw.py --arch SiamDW --resume snapshot/siamdw_res22w.pth --dataset VOT2017

Training

In root path $TracKit,

1. Download pretrain model from here and put it in pretrain (named with pretrain.model).

2. modify experiments/train/SiamDW.yaml according to your needs. (pls use GOT10K with 20w pairs each epoch in my opinion)

python tracking/train_siamdw.py

Then, pls follow the epoch testing and tuning as in Ocean.