cd SiamFC-Pytorch

mkdir models

# for color model
wget http://www.robots.ox.ac.uk/%7Eluca/stuff/siam-fc_nets/2016-08-17.net.mat -P models/
# for color+gray model
wget http://www.robots.ox.ac.uk/%7Eluca/stuff/siam-fc_nets/2016-08-17_gray025.net.mat -P models/

python bin/convert_pretrained_model.py

# video dir should conatin groundtruth_rect.txt which the same format like otb
python bin/demo_siamfc --gpu-id [gpu_id] --video-dir path/to/video

cd SiamFC-Pytorch

mkdir models

# using 12 threads should take an hour
python bin/create_dataset.py --data-dir path/to/data/ILSVRC2015 \
			     --output-dir path/to/data/ILSVRC_VID_CURATION \
			     --num-threads 8

# ILSVRC2015_VID_CURATION and ILSVRC2015_VID_CURATION.lmdb should be in the same directory
# the ILSVRC2015_VID_CURATION.lmdb should be about 34G or so
python bin/create_lmdb.py --data-dir path/to/data/ILSVRC_VID_CURATION \
			  --output-dir path/to/data/ILSVRC2015_VID_CURATION.lmdb \
		          --num-threads 8

# training should take about 1.5~2hrs on a Titan Xp GPU with 30 epochs
python bin/train_siamfc.py --gpu-id [gpu_id] --data-dir path/to/data/ILSVRC2015_VID_CURATION

Epoch 11 AUC: 0.5522
Epoch 12 AUC: 0.5670
Epoch 13 AUC: 0.5604
Epoch 14 AUC: 0.5559
Epoch 15 AUC: 0.5790
Epoch 16 AUC: 0.5687
Epoch 17 AUC: 0.5534
Epoch 18 AUC: 0.5745
Epoch 19 AUC: 0.5619
Epoch 20 AUC: 0.5749
Epoch 21 AUC: 0.5648
Epoch 22 AUC: 0.5775
Epoch 23 AUC: 0.5784
Epoch 24 AUC: 0.5812
Epoch 25 AUC: 0.5785
Epoch 26 AUC: 0.5637
Epoch 27 AUC: 0.5764
Epoch 28 AUC: 0.5675
Epoch 29 AUC: 0.5787
Epoch 30 AUC: 0.5820