Open data and code

Resources

Datasets and code released by the group.

Datasets

IN-SAT: land cover classification dataset for the Indian subcontinent

Sentinel-2 imagery for land cover classification. Published at IGARSS 2021 (paper).

In-Sat on GitHub

IIITD-20K: dense captioning for text-image person re-identification

20,000 identities captured in the wild, each image described in at least 26 words, with a benchmarking framework for text-to-image ReID models. Published at ICVGIP 2023 (paper).

Benchmark code on GitHub

Code

Code for most of our recent papers is on the group’s GitHub, and each paper on the publications page links to its own repository.