ORBIT: A Real-World Few-Shot Dataset for Teachable Object Recognition
D. Massiceti, L. Zintgraf, J. Bronskill, L. Theodorou, M. Harris, E. Cutrell, C. Morrison, K. Hofmann, S. Stumpf
This dataset is a collection of videos of objects recorded by people who are blind/low-vision on their mobile phones to drive research in Teachable Object Recognisers (TORs) under few-shot, high-variation conditions. Collectors recorded and submitted videos to the ORBIT benchmark dataset via an accessible iOS app.
Tags: blindnessblindLow vision