RC Community Datasets

Overview

There is a growing number of large public datasets that researchers rely on to conduct their work. Some of these datasets are utilized by different research groups or even different research fields, and as such, they are downloaded and hosted on the supercomputer in multiple file system locations.

To reduce the global load on the supercomputer’s shared filesystem and to foster data collaboration, we are pleased to consider hosting datasets in a shared location, specifically under /data/datasets.

Below is a table detailing the current datasets hosted for public use on the system. If you are interested in contributing to this community collection, please contact us with your request by reviewing our RTO Request Help page.

Current Community Datasets

Path

Short Description

Path

Short Description

Path

Short Description

Path

Short Description

/data/datasets/community/deeplearning/imagenet

Image database organized by the WordNet hierarchy http://www.image-net.org/

/data/datasets/community/compvis/LaSOT_categories

Large-scale Single Object Tracking (LaSOT) provides a collection of thousands of sequences with millions of frames across 70 categories. Each sequence is manually annotated http://vision.cs.stonybrook.edu/~lasot/

Benchmark results are given here: GitHub - HengLan/LaSOT_Evaluation_Toolkit: Evaluation of trackers on a large-scale benchmark LaSOT.

/data/datasets/community/compvis/OTB100

Data from the benchmark evaluation of online visual tracking algorithms (see: http://cvlab.hanyang.ac.kr/tracker_benchmark)

Raw zips and resulting images are hosted.

/data/datasets/community/compvis/TrackingNet

Work-in-progress

/data/datasets/community/blast_new

The complete BLAST databases, will be regularly updated.

Additional Help