/gazenet

A gazenet for mapping pupil position to gaze position based on resnet-18 & resnet-50

Primary LanguagePython

GazeNet

A gazenet for mapping pupil position to gaze position based on resnet-18 & resnet-50

Instructions

  • Train
python train_gazenet.py --datasets=data_train --saveas=model_name --log=log_name --test=True
  • Test
python run_gazenet.py --model=only_real.pt --test_data="test_data"

Tools

  • cal_err.py

    Calculate the error of world camera, magnified world camera and merged result and save the result as the format of .csv .

  • divide_datas&create_heatmap.py

    Divide the original data to train directory and test directory.

    Create the heatmap data based on the original coordinate data.

  • heatmap.py

    Create heatmap by coordinates.

  • sec_loc_to_main.py

    Convert the pupil position in magnified world camera to the position in world camera.

  • process_data.py

    Process the original data collected from Pupil-labs and classify them by users and points.

  • dsnt.py

    A network to inferring numerical coordinates for points of interest in an input image.

    [paper] [source code] [reference]

  • DSNT_example.py

    An example of DSNT.

  • covert_timestamps.py

    Covert the timestamps in 'gaze_positions.csv' collected from Pupil-labs.

Data

https://pan.baidu.com/s/1VrXuF1A1aFoKvd97LSZngw

  • data for train

    The training data is placed in the same level directory of the project.

    data_train:

    User1

    1_0.5667716914521796_0.3138741861260126

    test

    train

    2_0.5648374717682599_0.4591849427256318

    test

    train

    User2

    1_0.4707907267592169_0.49320440212082794

    test

    train

    2_0.6005824524164199_0.49661581171883473

    test

    train

  • data for test

    The test data is placed in the project directory.

    data_test:

    User34

    1_0.5977993905916809_0.31144010497464075

    2_0.5928179999323265_0.46354587733397024

    User35

    1_0.4549585918895902_0.4715157294162997

    2_0.5872032236896062_0.46782584350708645

Requirements