iSiam-TF

A TensorFlow implementation of the i-Siam tracker.

The codes were fetched and modified from https://github.com/bilylee/SiamFC-TensorFlow.

Introduction

This is a TensorFlow implementation of i-Siam: Improving Siamese Tracker with Distractors Suppression and Long-Term Strategies. If you use this code, please cite the following paper:

@inproceedings{TanISiam19,
  author    = {Tan, Wei Ren and Lai, Shang-Hong},
  title     = {i-Siam: Improving Siamese Tracker with Distractors Suppression and Long-Term Strategies},
  booktitle = {The IEEE International Conference on Computer Vision (ICCV) Workshops},
  year      = {2019},
}

Prerequisite

The main requirements can be installed by:

# (OPTIONAL) 0. It is highly recommended to create a virtualenv or conda environment
conda create -n pytf python=2.7
source activate pytf

# 1. Install TensorFlow
pip install tensorflow    # For CPU
pip install tensorflow-gpu  # For GPU

# 2. Install scipy for loading mat files
pip install scipy

# 3. Install sacred for experiments logging
pip install sacred

# 4. Install matplotlib for visualizing tracking results
pip install matplotlib

# 5. Install opencv for preprocessing training examples
pip install opencv-python

# 6. Install pillow for some image-related operations
pip install pillow

# 7. Install nvidia-ml-py for automatically selecting GPU
pip install nvidia-ml-py

# 8. Follow instructions in https://github.com/got-10k/toolkit to install GOT-10k toolkits.

Training

# 1. Download and unzip the GOT-10k dataset (http://got-10k.aitestunion.com/)
# Now, we assume it is unzipped to /path/to/got10k
DATASET=/path/to/got10k

# 2. Clone this repository to your disk 
# (Skip this step if you have already done)
git clone https://github.com/willtwr/iSiam-TF.git

# 3. Change working directory
cd iSiam-TF

# 4. Create a soft link pointing to the GOT-10k dataset
mkdir -p data
ln -s $DATASET data/got10k

# 5. Prepare training data
python scripts/preprocess_got10k_data.py

# 6. Split train/val dataset and store corresponding image paths
python scripts/build_got10k_imdb_reg.py

# 7. Start training
python experiments/iSiam.py

# 8. (OPTIONAL) View the training progress in TensorBoard
# Open a new terminal session and cd to iSiam-TF, then
tensorboard --logdir=Logs/SiamFC/track_model_checkpoints/iSiam

Benchmark OTB-100

Benchmark for OTB-100 uses the custom OTB evaluation toolkit where several bugs are fixed.

# Assume directory structure:
# Your-Workspace-Directory
#         |- iSiam-TF
#         |- tracker_benchmark
#         |- ...
# 0. Go to your workspace directory
cd /path/to/Your-Workspace-Directory

# 1. Download the OTB toolkit
git clone https://github.com/bilylee/tracker_benchmark.git

# 2. Modify iSiam-TF/benchmarks/run_iSiam_otb.py if needed. 

# 3. Copy run_iSiam_otb.py to the evaluation toolkit
cp iSiam-TF/benchmarks/run_iSiam_otb.py tracker_benchmark/scripts/bscripts

# 4. Add the tracker to the evaluation toolkit list
echo "\nfrom run_iSiam_otb import *" >> tracker_benchmark/scripts/bscripts/__init__.py

# 5. Create tracker directory in the evaluation toolkit
mkdir tracker_benchmark/trackers/iSiam_otb

# 6. Start evaluation (it will take some time to download test sequences).
echo "tb100" | python tracker_benchmark/run_trackers.py -t iSiam_otb -s tb100 -e OPE

# 7. Get the AUC score
sed -i "s+tb50+tb100+g" tracker_benchmark/draw_graph.py
python tracker_benchmark/draw_graph.py

Benchmark UAV-123

Benchmark for UAV-123.

# Assume directory structure:
# Your-Workspace-Directory
#         |- iSiam-TF
#         |- ...
# 0. Go to your workspace directory
cd /path/to/Your-Workspace-Directory/iSiam-TF

# 1. Modify uav123.py if needed, e.g. all the paths. 

# 2. Start evaluation (it will take some time to download test sequences).
python uav123.py

# (optional) 3. Get the AUC score for all compared trackers
python uav123_draw.py

Benchmark TLP

Benchmark for TLP uses another custom toolkit.

# Assume directory structure:
# Your-Workspace-Directory
#         |- iSiam-TF
#         |- TLP_benchmark
#         |- ...
# 0. Go to your workspace directory
cd /path/to/Your-Workspace-Directory

# 1. Download the custom TLP toolkit
git clone https://github.com/willtwr/TLP_benchmark.git

# 2. Modify TLP_benchmark/config.py if needed. 

# 3. Modify TLP_benchmark/scripts/bscripts/run_iSiam.py if needed. 

# 4. Add the tracker to the evaluation toolkit list
echo "\nfrom run_iSiam import *" >> TLP_benchmark/scripts/bscripts/__init__.py

# 5. Create tracker directory in the evaluation toolkit
mkdir TLP_benchmark/trackers/iSiam

# 6. Start evaluation (it will take some time to download test sequences).
echo "tlp" | python tracker_benchmark/run_trackers.py -t iSiam -s tlp -e OPE

# 7. Get the AUC score
sed -i "s+tb50+tb100+g" TLP_benchmark/draw_graph.py
python TLP_benchmark/draw_graph.py

Benchmark OxUvA

Benchmark for OxUvA. Please follow the instructions in OxUvA github for installation. Note that users need to modify the paths accordingly, both in the command line and .py files.

# Assume directory structure:
# Your-Workspace-Directory
#         |- iSiam-TF
#         |- long-term-tracking-benchmark
#         |- ...
# 0. copy iSiam-TF/benchmarks/oxuva/scripts to long-term-tracking-benchmark/

# 1. copy iSiam-TF/benchmarks/oxuva/examples to long-term-tracking-benchmark/examples

# 2. Go to 
cd /path/to/Your-Workspace-Directory/long-term-tracking-benchmark/examples

# 3. Start evaluation (it will take some time to download test sequences).
python track.py -v ../../dataset/ ../../predictions/ --data=dev --tracker=iSiam

# 4. To draw graphs, please follow the instructions in https://github.com/oxuva/long-term-tracking-benchmark.git

License

iSiam-TF is released under the MIT License (refer to the LICENSE file for details).

willtwr / iSiam-TF