Algorithms, computation and visual data are the three pillars of computer vision (CV). Researchers, institutions and open source communities have produced sophisticated algorithms and open-sourced code; while global tech giants’ supercharged cloud platforms provide all the computational power CV researchers require. However, efficiently sourcing visual data — particularly images with high-quality annotations — remains a challenge.
Building large datasets is a time-consuming and labor-intensive task which challenges entities with limited budgets. There are hundreds of open visual datasets out there, but searching across them and their millions of entries is not a simple task.
VisualData is a search engine for image datasets created by Jie Feng, a Columbia University PhD graduate. Feng launched the project to “help people interested in computer vision easily explore existing data to do their experiments faster.” The VisualData website allows users to search and download from image datasets contributed by researchers, hobbyists, and businesses.
As of June 2019 the VisualData website contains 295 labeled datasets, among them the YouTube-VOS Large-Scale Benchmark for Video Object Segmentation Dataset; Tencent ML-Images, an 18 million multi-label image dataset with 11,000 categories; and Flickr-Faces-HQ Flickr-Faces-HQ, a high-quality image dataset of human faces. There are also many highly specialized datasets. Image classes range from 3D reconstructions and faces to wild animals, robots, fashion and more.
The datasets can be sorted by published date or topic, and users can search with keywords to locate images appropriate to their needs. The VisualData website also shows each dataset’s code availability and popularity. Users can easily contribute their own datasets via an “Add My Dataset” button; and can register to receive email updates when new datasets are added to the project.
Feng works as an applied scientist at Amazon, where his primary interests are computer vision and machine learning.
For more information, check the VisualData website.
Author: Hongxi Li | Editor: Michael Sarazen; Tony Peng