A Solution to Co-occurrence Bias: Attributes Disentanglement via Mutual Information Minimization for Pedestrian Attribute Recognition
This addresses a domain-specific problem for pedestrian attribute recognition in computer vision, improving robustness in realistic scenes by reducing bias, but it is incremental as it builds on existing methods for disentanglement.
The paper tackles the problem of co-occurrence bias in pedestrian attribute recognition, where models overfit to attribute interdependencies in training data, by proposing an attributes-disentangled feature learning method that minimizes mutual information to decouple attributes, achieving state-of-the-art performance on realistic datasets like PETAzs and RAPzs.
Recent studies on pedestrian attribute recognition progress with either explicit or implicit modeling of the co-occurrence among attributes. Considering that this known a prior is highly variable and unforeseeable regarding the specific scenarios, we show that current methods can actually suffer in generalizing such fitted attributes interdependencies onto scenes or identities off the dataset distribution, resulting in the underlined bias of attributes co-occurrence. To render models robust in realistic scenes, we propose the attributes-disentangled feature learning to ensure the recognition of an attribute not inferring on the existence of others, and which is sequentially formulated as a problem of mutual information minimization. Rooting from it, practical strategies are devised to efficiently decouple attributes, which substantially improve the baseline and establish state-of-the-art performance on realistic datasets like PETAzs and RAPzs. Code is released on https://github.com/SDret/A-Solution-to-Co-occurence-Bias-in-Pedestrian-Attribute-Recognition.