1996scarlet
commented
4 years ago Well, since the author of that paper is not me. When reproducing his model, I found that the effect of adding a FC layer to classify the eye as close and open was very bad, so I deleted this structure. I judge the state of the eye by counting the number of iris pixels.
Gautam Malu notifications@github.com 于2020年4月22日周三 下午5:11写道：
@1996scarlet https://github.com/1996scarlet In the paper "Realtime and Accurate 3D Eye Gaze Capture with DCNN-based Iris and Pupil Segmentation" you have mentioned a separate FC branch in IRIS segmentation model to classify the eye as close and open. I am not able to find that branch in your shared model here https://github.com/1996scarlet/Laser-Eye/blob/master/weights/iris-symbol.json
— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/1996scarlet/Laser-Eye/issues/7, or unsubscribe https://github.com/notifications/unsubscribe-auth/AFO2D4JSZZTLOLR6FJ5TX53RN2YFRANCNFSM4MN7RJSA .
@1996scarlet In the paper "Realtime and Accurate 3D Eye Gaze Capture with DCNN-based Iris and Pupil Segmentation" you have mentioned a separate FC branch in IRIS segmentation model to classify the eye as close and open. I am not able to find that branch in your shared model here https://github.com/1996scarlet/Laser-Eye/blob/master/weights/iris-symbol.json