yuezunli / wifs2018_in_ictu_oculi Goto Github PK

I prepare 40 standarad eye images, When I use the model to predict , I found that the same eye image got very different prob. Wonder Why it happened。
eg.
[[0.9415097 0.05849024]]
and
[[0.35040304 0.6495969 ]]

Hello authors, Your work is fantastic. My issue is when I pass a video with 1920*1080 px(captured by my mobile(around 10 mb size)), the face detecor part shows unpredictable behevious. It gives None ....
face_cache = lib.align(im[:, :, (2,1,0)], self.front_face_detector, self.lmark_predictor)

But when I compress my video(3-4mb) by using any online video compression tool its works fine. Can you help me with it.

It is so amazing. Another paper related about head pose estimation detection test on this dataset. Maybe some authors involve this work too. Your dataset is not released and your code test on this work is not released.
Such an amazing work!

Hello,

sample_eye_data/train.p is not generated. Could you please clarify which script generates train.p file.

Thank you.

Hey,
I have collected sequence of image data as specified.
Kindly can you give details about the format of annotated file required for LRCN training.
Thank you,

Hi,
This project looks interesting. I wanted to know the dataset used for training.

Thank you,

so amazing your project, I am a phd stu in NUS, may I know the correct website address for you dataset? It seems the one in paper can not be found.