Open cccusername opened 4 years ago
@cccusername We randomly sampled 700 classes from ImageNet as described in paper (animals (15%), plants (15%), people (20%), scenery (25%), food (5%), transportation (15%) and artifacts (5%)). For each class, we randomly 1,000 sample image pairs with different extents of similarity (45% of Top-5 similarity searched used the code of Gray-Image-Retrieval, 45% of randomly sampled within the same class, and 10% of randomly sampled within the same category). The major part is the image retrieval as we shared in this repo.
hello, @hmmlillian . If possible, could your share the code to creaate the dataset? Thank you!