Multi-label Classification with Partial Annotations using Class-aware Selective Loss

TL;DR

This paper introduces a novel class-aware selective loss method for large-scale multi-label classification with partial annotations, improving label handling by estimating class distributions and emphasizing annotated labels, leading to state-of-the-art results.

Contribution

The paper proposes a new approach combining class distribution estimation and asymmetric loss to better handle partial labels in multi-label classification.

Findings

Achieved 87.3 mAP on OpenImages V6 dataset.

Outperformed existing methods on LVIS and simulated-COCO datasets.

Demonstrated the effectiveness of class-aware selective loss in partial annotation scenarios.

Abstract

Large-scale multi-label classification datasets are commonly, and perhaps inevitably, partially annotated. That is, only a small subset of labels are annotated per sample. Different methods for handling the missing labels induce different properties on the model and impact its accuracy. In this work, we analyze the partial labeling problem, then propose a solution based on two key ideas. First, un-annotated labels should be treated selectively according to two probability quantities: the class distribution in the overall dataset and the specific label likelihood for a given data sample. We propose to estimate the class distribution using a dedicated temporary model, and we show its improved efficiency over a naive estimation computed using the dataset's partial annotations. Second, during the training of the target model, we emphasize the contribution of annotated labels over originally un-annotated labels by using a dedicated asymmetric loss. With our novel approach, we achieve state-of-the-art results on OpenImages dataset (e.g. reaching 87.3 mAP on V6). In addition, experiments conducted on LVIS and simulated-COCO demonstrate the effectiveness of our approach. Code is available at https://github.com/Alibaba-MIIL/PartialLabelingCSL.