search

nnizhang / S2MA

source code of "Learning Selective Self-Mutual Attention for RGB-D Saliency Detection" (CVPR2020)
62 stars 8 forks source link

NJU2k missing 100 samples? #3

Open fereenwong opened 3 years ago

fereenwong commented 3 years ago

hi, thanks for sharing the great work. I download the provided dataset, RGBdDataset_processed. In which, the NJU2K dataset contains 1400/485 samples in trainset/testset folds, 1885 samples in total. Are there missing 100 samples? (As noted in the paper, NJUD includes 1985 samples.)

nnizhang commented 3 years ago

For fair comparisons, we adopt the same trainset and test set as in [1] for the NJU2K dataset. As described in [1], the authors sample 100 images for the validation set and 1400/485 for the trainset/test set from the NJU2K dataset.

[1] Hao Chen and Youfu Li. Progressively complementarity-aware fusion network for rgb-d salient object detection. In CVPR, pages 3051–3060, 2018.