Adaptive Binarization for Weakly Supervised Affordance Segmentation

TL;DR

This paper introduces an adaptive binarization method for weakly supervised affordance segmentation, improving the learning process of CNNs from sparse keypoints and outperforming existing methods on benchmark datasets.

Contribution

It proposes an adaptive binarization technique with parameter estimation via approximated cross validation for better weakly supervised affordance segmentation.

Findings

Outperforms state-of-the-art in weakly supervised affordance segmentation

Effective adaptive binarization improves CNN training from sparse keypoints

Validated on two benchmark datasets

Abstract

The concept of affordance is important to understand the relevance of object parts for a certain functional interaction. Affordance types generalize across object categories and are not mutually exclusive. This makes the segmentation of affordance regions of objects in images a difficult task. In this work, we build on an iterative approach that learns a convolutional neural network for affordance segmentation from sparse keypoints. During this process, the predictions of the network need to be binarized. In this work, we propose an adaptive approach for binarization and estimate the parameters for initialization by approximated cross validation. We evaluate our approach on two affordance datasets where our approach outperforms the state-of-the-art for weakly supervised affordance segmentation.