Similarity

Demos hosted on this machine.

Near Duplicates demo
Poly Query Demo
The Mirflkr images are all hosted here at adresses such as https://similarity.cs.st-andrews.ac.uk/mirflickr/images/42/429988.jpg in directories containing 10,000 images each hence the 42,0000 - 42,9999 images are all the same directory. Thus image 429988.jpg is:
a flower
The thumb nail images are also on this server, the thumb for the above flower may be found at the address: https://similarity.cs.st-andrews.ac.uk/mirflickr/thumbs/43/im429989.jpg, note that these are indexed from 1 not zero!
a flower
The nearest neighbours are also here. The near neighbours are encoded in 1000 files: 0.txt - 999.txt with the NN data
Each file is in the format:
0 0 0.0 585585 252.89404491750705 155784 848.0200759945012 733551 890.244332062840…
line ::= <id> <tab> [ <id> <tab> <dist> ]* \n <br> <dist> ::= <ascii double>
Thus the first of these is at the address https://similarity.cs.st-andrews.ac.uk/mirflickr/NN_gt/1.txt.

Places365 Source Images

This server contains the training image set of the Places365 dataset. The URL schema used to access these images is:
https://similarity.cs.st-andrews.ac.uk/p365/train/< category>/< image_id>.jpg

Most categories contain 5,000 images, although some contain less. These are numbered 00000001 to 00005000.
The full index of categories can be found here.
The full index of sizes (ordered by category as above) can be found here.
See random images here This template locates all images.
Image category:
Image count (type 1-5000):