DeepFH Segmentations for Superpixel-based Object Proposal Refinement

TL;DR

This paper introduces a superpixel-based refinement system using DeepFH segmentation to improve the accuracy of object proposals in detection pipelines, outperforming existing methods on COCO with LVIS annotations.

Contribution

The paper presents a novel DeepFH segmentation method that combines deep features with Felzenszwalb and Huttenlocher segmentation for enhanced object proposal refinement.

Findings

DeepFH segmentation improves segmentation quality.

Refinement with DeepFH outperforms state-of-the-art methods.

Enhanced proposals lead to better object detection accuracy.

Abstract

Class-agnostic object proposal generation is an important first step in many object detection pipelines. However, object proposals of modern systems are rather inaccurate in terms of segmentation and only roughly adhere to object boundaries. Since typical refinement steps are usually not applicable to thousands of proposals, we propose a superpixel-based refinement system for object proposal generation systems. Utilizing precise superpixels and superpixel pooling on deep features, we refine initial coarse proposals in an end-to-end learned system. Furthermore, we propose a novel DeepFH segmentation, which enriches the classic Felzenszwalb and Huttenlocher (FH) segmentation with deep features leading to improved segmentation results and better object proposal refinements. On the COCO dataset with LVIS annotations, we show that our refinement based on DeepFH superpixels outperforms state-of-the-art methods and leads to more precise object proposals.