search

chenkang455 / TRMD

[TMM 2024] Motion Deblur by Learning Residual from Events
13 stars 0 forks source link

Question #7

Open JIEyuUUU0818 opened 2 weeks ago

JIEyuUUU0818 commented 2 weeks ago

Hello author, I would like to ask if the results compared in Table 1 of the paper, "QUANTIVE Results of SINGLE IMAGE DEBLRING," are using the dataset you provided or the datasets used by these papers themselves.

chenkang455 commented 2 weeks ago

Hi, @JIEyuUUU0818, since our method requires a sequence of sharp images for supervision, existing Event-GOPRO datasets are not suitable for our task during the experiments period. Methods with source code like EFNet are re-trained on our dataset, while methods without the source code are evaluated directly on our dataset.

JIEyuUUU0818 commented 1 week ago

Thank you for your reply, author. Recently, I have been trying to run the EFNet model using your dataset. I would like to ask how to train the EFNet model using your own constructed dataset, and which parts need to be modified to enable the EFNet model to successfully read the dataset and train. I would greatly appreciate your reply.