Dataset Details (DDN-SIRR)

[kylin013]

Dataset Details (DDN-SIRR)

[rajeevyasarla]

Hi @kylin013 ,

1. yes, we used 147 real-world images in the training set as unlabelled data set, and they are provided in "./data/train/derain/real_input/"
2. Yes, the original dataset of Fu et al. [9] consists of 9100 training pairs and 4900 testing pairs, in this publication we didn't use 4900 pairs for testing.
3. updated dataset of Fu et al. [9] consists of 12600:1400 training: testing split. their dataset is created using 14 rain streaks for each image, so using 1000 images a total of 14000 images were created and split into training and testing. Earlier I think images index with 1 to 650 were used for training and rest for testing. In the updated dataset images indexed with 1to 900 are used as training and 901 to 1000 are for testing. So in our training processes we used images indexed with 1to 650 as labelled dataset in our training.

[kylin013]

Hi @rajeevyasarla , thanks so much for your reply! I still have a question, did you use the 20 images in "./data/test/derain/SIRR_test/" for both validation and testing, and which 10 of them are dense (or sparse)?

[rajeevyasarla]

Hi @kylin013 , Yes I use 20 images "./data/test/derain/SIRR_test/" for testing where 10 are dense and 10 are sparse

[kylin013]

Hi @rajeevyasarla , which 10 images are dense? the images index with 1 to 10?

[rajeevyasarla]

Hi @kylin013 , images indexed from 1 to 10 are dense, and 11-20 are sparse