search

phoenix104104 / fast_blind_video_consistency

Learning Blind Video Temporal Consistency (ECCV 2018)
http://vllab.ucmerced.edu/wlai24/video_consistency/
Other
423 stars 64 forks source link

wrong input #2

Closed Gogumee closed 6 years ago

Gogumee commented 6 years ago

Hi!

I think your model's input is different from your paper.

[Code]

model input

inputs = torch.cat((frame_p2, frame_o1, frame_i2, frame_o1), dim=1)

should be inputs = torch.cat((frame_p2, frame_i1, frame_i2, frame_o1), dim=1)

is it right?

Thank you for your good research :)

phoenix104104 commented 6 years ago

Hi, the order of o1, i1, and i2 doesn't matter if it's consistency during training and testing.

musicrainie commented 6 years ago

Hi!

I think your model's input is different from your paper.

[Code]

model input

inputs = torch.cat((frame_p2, frame_o1, frame_i2, frame_o1), dim=1)

should be inputs = torch.cat((frame_p2, frame_i1, frame_i2, frame_o1), dim=1)

is it right?

Thank you for your good research :)

According to the test_pretrained.py and the provided results.zip, inputs to the model should be (frame_p2, frame_o1, frame_i2, frame_o1), which surely seems not to comply with the paper.