Authors

• Sanjeev Kumar Singh
• Ankur Kunder

Results on vot2013/bicycle:

Models	FPS
ADNet	2.90
ADNet-Fast	15.00
Ours	9.65

Python version: This code is in Python3.6

Package Requirements: tensorflow==1.14.0, OpenCV==3.4.2, pyyaml, hdf5storage, h5py

Graphs

• The plots shown are the obsrvations collected on single training video

Our Model

Implementation of 'Action-Decision Networks for Visual Tracking with Deep Reinforcement Learning(CVPR 2017)'

GIF	Description
	vot2013 Dataset bicycle

• Green : Ground Truth, Blue : Ours implemented model

Data Preperation:

Step 1

Download the Visual Object Tracking Dataset from website. We have used the data for 2013, 2014 and 2015. More data always helps.

Step 2

Put the inside folder train_data as vot2013, vot2014 and vot2014. This folder should have folder video wise and inside each folder. extract frames in img folder.

Model Training

python runner.py -debug <bool denoting debug on 1 video> -mode <train|test> -rl <bool denoting if RL is on> -model_path <model-path> -vid_path <video-path>

• Loss will be logged at model_path/loss.log
• You can download our pre-trained model at this link.

Acknowledgements

Base Code is take from here