perhuaper / siamfc_got Goto Github PK

modified from (https://github.com/got-10k/siamfc).

add test-and-record while training

add multiprocess to test.py

modified show_image() to visualize

support changing backbone to vgg (siamvgg)

install pytorch, opencv-python:

pip install torch opencv-python

1. Download pretrained model_siamfc.pth from Baidu Yun. sqrb

2. Download pretrained model_siamvgg.pth from Baidu Yun. vskn

3. Download GOT10k dataset and OTB2015 dataset.

4. Change your datasets folder dir and model dir (in test.py):

5. Run:

python test.py --tracker siamfc

python test.py --tracker siamvgg

1. Change dir to your datasets folder (in end of train.py):

2. Run:

python train.py --tracker siamfc

python train.py --tracker siamvgg

By default, tracker will train in GOT10k and test over OTB2015 dataset.

####combine epoch15 of SiamFC and epoch4 of SiamVGG (only test)