vimukthipahasaragodage / dvgaze Goto Github PK

DVGaze: Dual-view Gaze Estimation

Yihua Cheng and Feng Lu, ICCV 2023

Dual cameras have been applied in many devices recently. In this paper, we explore a new direction for gaze estimation. We propose a dual-view gaze estimation network (DV-Gaze) including dual-view interactive convolution block and dual-view transformers.

Please re-write the config file in config/train/xxx.yaml and config/test/xxx.yaml.

To train a model, Please run the command:

python trainer/total.py config/train/xxx.yaml

To evaluate a model, please run the command:

python tester/total.py config/train/xxx.yaml config/test/xxx.yaml

where config/train/xxx.yaml indicates the config for training model.

Please download the label file from Google Driver . We are not authorized to distribute image files. Please access Gazehub to acquire data-processing code on ETH and EVE. Note that the website only works on working time (8:00-19:00?) in the China time zone (I have graduated and don't know the reason). Apologise for the inconvenience. I will try my best to update data-processing codes in Github.

All codes are beta version and we will keep updating this repositories. The full version will be updated before Oct. 10.

Please send email to [email protected] if you have any questions.

We can also meet in ICCV conference in person:) Have a good day.