This project allows to create Web Interface for recording the gaze on videos. We used this interface for recording gaze by crowdsourcing the task in Amazon Mechanical Turk for our following work:

"Object Referring in Videos using Language and Human Gaze" - CVPR 2018
Arun Balajee Vasudevan, Dengxin Dai, Luc Van Gool
CVLab, ETH Zurich

1. Download this repository or clone with Git, then cd into the root directory.
2. Copy the folder where your host server have access to. And, mention the path of the folder in L5 in .htaccess
3. Define your python packages folder in L8 in gaze_int.fcgi. (create the package locally visible to the server)
4. Install the requirements with pip install -r requirements.txt
5. Run FastCGI as ./gaze_int.fcgi in the host server. For debugging, check server.log.

User input to the interface: application/boxes.csv.

Check demo at Firefox browser:

https://people.ee.ethz.ch/~arunv/gaze-interface/gaze_int.fcgi/?image_id=zurich_000005_000029_leftImg8bit