Dataset of our ICIP2022 paper: "Semi-Supervised Ranking for Object Image Blur Assessment".
We realease the txt files on the "Release" of this repo: https://github.com/yzliangHIK2022/SSRanking-for-Object-BA/releases/tag/releaseDataset_1.0
Please kindly refer to above URL for downloading.
We provide the face cropped information of dataset images in:
LFW_all_imgs_list.txt, CASIA_all_imgs_list.txt, WiderFace_train_all_imgs_list.txt, WiderFace_val_all_imgs_list.txt, MegaFace_all_imgs_list.txt.
The format (for each line) of cropped txtfiles:
"DatasetName>>Original_image_name>>crop_x crop_y crop_width crop_height>>Image_name_after_cropping\n".
e.g. "LFW>>Aaron_Eckhart_0001.jpg>>22 33 202 202>>Aaron_Eckhart_0001.jpg\n": It means that image "Aaron_Eckhart_0001.jpg" is cropped from "LFW" dataset based on position of "22 33 202 202" in original image "Aaron_Eckhart_0001.jpg".
We release our supervised and unsupervised tranning list:
SSRanking_OBA_train_supervised_5k.txt, SSRanking_OBA_train_supervised_10k.txt, SSRanking_OBA_train_unsupervised.txt.
The format (for each line) of supervised txtfiles:
"Image_left,Image_right\n".
e.g. "CASIA_WebFace/0000133/039.jpg,LFW/Chris_Neil_0001.jpg\n": It means that the face in the "Image_right" in more blur than "Image_left".
The format (for each line) of unsupervised txtfiles:
"Image_left,Image_right\n".
e.g. "Images_Wider_Train/35--Basketball/35_Basketball_playingbasketball_35_397_2.jpg,Images_Wider_Train/0--Parade/0_Parade_Parade_0_69_5.jpg\n": Image "Image_left" and image "Image_right" will be used as unsupervised image-pair for unsupervised training by L_{QRC} which is detailed in our paper.
We release our testing list:
SSRanking_OBA_test_test1.txt, SSRanking_OBA_test_test2.txt SSRanking_OBA_test_test3.txt.
The format (for each line) of txtfiles:
"image_name image_order\n".
e.g. "LFW/Gordon_Brown_0007_2.jpg 1\n": It means that image "image_name" rank order "image_order" in current test set. The larger the rank order, the more blur is this image.
@misc{2207.06085,
Author = {Qiang Li and Zhaoliang Yao and Jingjing Wang and Ye Tian and Pengju Yang and Di Xie and Shiliang Pu},
Title = {Semi-supervised Ranking for Object Image Blur Assessment},
Year = {2022},
Eprint = {arXiv:2207.06085},
}