WSSOD: A New Pipeline for Weakly- and Semi-Supervised Object Detection

TL;DR

This paper introduces WSSOD, a novel weakly- and semi-supervised object detection framework that effectively balances detection accuracy and annotation cost by leveraging both fully and weakly labeled data through a two-stage learning process and innovative strategies.

Contribution

The paper proposes a new WSSOD framework with a two-stage learning process, incorporating weakly-supervised loss, label attention, and pseudo-label sampling to improve detection with limited annotations.

Findings

Achieves high detection performance with only one third of annotations.

Demonstrates effectiveness on PASCAL-VOC and MSCOCO datasets.

Comparable to fully-supervised methods in accuracy.

Abstract

The performance of object detection, to a great extent, depends on the availability of large annotated datasets. To alleviate the annotation cost, the research community has explored a number of ways to exploit unlabeled or weakly labeled data. However, such efforts have met with limited success so far. In this work, we revisit the problem with a pragmatic standpoint, trying to explore a new balance between detection performance and annotation cost by jointly exploiting fully and weakly annotated data. Specifically, we propose a weakly- and semi-supervised object detection framework (WSSOD), which involves a two-stage learning procedure. An agent detector is first trained on a joint dataset and then used to predict pseudo bounding boxes on weakly-annotated images. The underlying assumptions in the current as well as common semi-supervised pipelines are also carefully examined under a unified EM formulation. On top of this framework, weakly-supervised loss (WSL), label attention and random pseudo-label sampling (RPS) strategies are introduced to relax these assumptions, bringing additional improvement on the efficacy of the detection pipeline. The proposed framework demonstrates remarkable performance on PASCAL-VOC and MSCOCO benchmark, achieving a high performance comparable to those obtained in fully-supervised settings, with only one third of the annotations.