Python (Tensorflow) implementation of CFNet tracker
- Get virtualenv if you don't have it already
pip install virtualenv
- Create new virtualenv with Python 2.7
virtualenv --python=/usr/bin/python2.7 ve-tracking
- Activate the virtualenv
source ~/ve-tracking/bin/activate
- Clone the repository
https://github.com/Wenju-Huang/py-CFNet.git
cd py-CFNet
- Install the required packages
sudo pip install -r requirements.txt
mkdir pretrained data
- Download the pretrained networks in
pretrained
and unzip the archive
- Set
root_dataset
inparameters/environment.json
to the root path of your evaluation dataset (don't include the dataset name) - Set
dataset
inparameters/evaluation.json
to the name your evaluation dataset (e.g."CVPR2013"
) - Set
video
inparameters/evaluation.json
to"all"
or to a specific sequence (e.g."vot2016_ball1"
) - See if you are happy with the default parameters in
parameters/hyperparameters.json
- Select the model you want to run at line 18 in
python run_tracker_evaluation.py
(default is'conv2'
) - Call the main script (within an active virtualenv session)
python run_tracker_evaluation.py
CFNetet-Conv2--OTB2013: Precision (20 px): 73.83 -- Success AUC: 0.55 -- FPS: 42.97 FPS
Many parts of this code are adopted from the related works siamfc-tf