Data pre-processing

[Ijustakid]

I am using a real mobile device to capture the data.

Using opencv for greyscale processing and normalisation, the image is very different from pre_img.png.

My grayed-out image, which is not as detailed as the one you provided. For example, pupil detail is completely lost.

I would like to ask if there is anything special about the image pre-processing of this dataset.Or is there any way to keep my captured images consistent with the training set.

BR.

[ycdhqzhiai]

demo里面的图片没什么特别的处理，就是数据集里面的一张，你可以参考下paper，摄像头 好像是一种头戴设备专门采集眼睛区域用来做VR的。