Ruicong Liu

icon indicating a website link https://ruicongliu.github.io/ icon indicating an email [email protected]

icon company UTokyo icon location Tokyo icon location -

Results 5 comments of Ruicong Liu

AttributeError: module 'mediapipe.python.solutions.face_mesh' has no attribute 'FACEMESH_CONTOURS'

I replaced 'FACEMESH_CONTOURS' with 'FACE_CONNECTIONS'

image size

Hi, it depends on the image size while pre-training. The size should keep the same between pre-training and adaptation.

Pretrained models

Thanks, the pre-models are: https://drive.google.com/file/d/1qgecpBK6oFkcoyqpBO4vTb7P3Rx7Z--m/view?usp=sharing

target和source的图像数量不一致

你好，adapt,py 的'--pics'参数在100，102行输入txtload函数，该参数保证了读入图片的数量。 其次，zip(dataset_target, dataset_source)也可以保证数量一致^_^。

Something about data

The label conversion follows "MPIIGaze: Real-World Dataset and Deep Appearance-Based Gaze Estimation". Things like "1_A_DS_S_1" are from the raw data. Only "face" and "gaze2d" are used during adaptation.