On Label Granularity and Object Localization
This addresses a practical problem in WSOL for computer vision researchers, offering an incremental improvement by optimizing label usage rather than developing new methods.
The paper investigates how label granularity affects weakly supervised object localization (WSOL), finding that selecting appropriate training labels yields a larger performance boost than algorithm choice and improves data efficiency.
Weakly supervised object localization (WSOL) aims to learn representations that encode object location using only image-level category labels. However, many objects can be labeled at different levels of granularity. Is it an animal, a bird, or a great horned owl? Which image-level labels should we use? In this paper we study the role of label granularity in WSOL. To facilitate this investigation we introduce iNatLoc500, a new large-scale fine-grained benchmark dataset for WSOL. Surprisingly, we find that choosing the right training label granularity provides a much larger performance boost than choosing the best WSOL algorithm. We also show that changing the label granularity can significantly improve data efficiency.