Visual Tracking with Fully Convolutional Networks (FCNT)

Overview

This project is a Python reimplementation based on the ICCV 2015 paper Visual Tracking with Fully Convolutional Networks by Lijun Wang, Wanli Ouyang, Xiaogang Wang and Huchuan Lu.

Official Implementation

The Official MATLAB implementation can be accessed via the paper's GitHub repository.

---

🛠️ Configuration

Before running the tracker, you must edit the config file:

configs/config.yaml

In particular, change these values:

sequence_path: data/sample_sequence/      # Path to your video sequence
init_bbox: [x, y, w, h]                   # Initial bounding box in the first frame

🔧 Environment Setup

This project uses Conda. Run the script below to create the environment and run the code after configuring config.yaml or you can use the command in section Run Tracker:

bash run_fcnt.sh

🚀 Run Tracker

Once the config file is set, you can also run the tracker as follow:

python run.py --config configs/config.yaml