StuffNet: Using 'Stuff' to Improve Object Detection

TL;DR

StuffNet enhances object detection by integrating 'stuff' segmentation features, significantly improving performance especially for small objects, and can be trained on datasets lacking explicit 'stuff' labels.

Contribution

The paper introduces StuffNet, a CNN that incorporates 'stuff' segmentation features to improve object detection, including a training method for datasets without 'stuff' labels.

Findings

Improves small object detection from 18.8% to 23.9% mAP.

Effective on datasets without 'stuff' segmentation labels.

Significantly boosts overall object detection performance.

Abstract

We propose a Convolutional Neural Network (CNN) based algorithm - StuffNet - for object detection. In addition to the standard convolutional features trained for region proposal and object detection [31], StuffNet uses convolutional features trained for segmentation of objects and 'stuff' (amorphous categories such as ground and water). Through experiments on Pascal VOC 2010, we show the importance of features learnt from stuff segmentation for improving object detection performance. StuffNet improves performance from 18.8% mAP to 23.9% mAP for small objects. We also devise a method to train StuffNet on datasets that do not have stuff segmentation labels. Through experiments on Pascal VOC 2007 and 2012, we demonstrate the effectiveness of this method and show that StuffNet also significantly improves object detection performance on such datasets.