segDeepM: Exploiting Segmentation and Context in Deep Neural Networks for Object Detection

TL;DR

This paper introduces segDeepM, a method that enhances object detection accuracy by integrating segmentation and contextual cues within a Markov Random Field framework, leveraging CNNs to select and score segmentation proposals.

Contribution

It presents a novel approach combining segmentation and context in deep neural networks for improved object detection accuracy.

Findings

Achieved 4.1% improvement in mAP over R-CNN on PASCAL VOC 2010

Surpassed current state-of-the-art by 3.4% in mAP

Demonstrated the effectiveness of segmentation and context integration

Abstract

In this paper, we propose an approach that exploits object segmentation in order to improve the accuracy of object detection. We frame the problem as inference in a Markov Random Field, in which each detection hypothesis scores object appearance as well as contextual information using Convolutional Neural Networks, and allows the hypothesis to choose and score a segment out of a large pool of accurate object segmentation proposals. This enables the detector to incorporate additional evidence when it is available and thus results in more accurate detections. Our experiments show an improvement of 4.1% in mAP over the R-CNN baseline on PASCAL VOC 2010, and 3.4% over the current state-of-the-art, demonstrating the power of our approach.