G-RCN: Optimizing the Gap between Classification and Localization Tasks for Object Detection

TL;DR

This paper introduces G-RCN, a novel object detection paradigm that separates classification and localization tasks to optimize their performance, leading to significant improvements on standard datasets with minimal structural changes.

Contribution

The paper proposes G-RCN, a new multi-task learning approach that separates classification and localization tasks, improving accuracy without adding extra modules.

Findings

G-RCN improves AP70 by 3.6 on PASCAL VOC with ResNet50.

G-RCN enhances AP by 1.5 on COCO with ResNet50.

Applying G-RCN to various backbones yields over 2.0 AP70 improvement.

Abstract

Multi-task learning is widely used in computer vision. Currently, object detection models utilize shared feature map to complete classification and localization tasks simultaneously. By comparing the performance between the original Faster R-CNN and that with partially separated feature maps, we show that: (1) Sharing high-level features for the classification and localization tasks is sub-optimal; (2) Large stride is beneficial for classification but harmful for localization; (3) Global context information could improve the performance of classification. Based on these findings, we proposed a paradigm called Gap-optimized region based convolutional network (G-RCN), which aims to separating these two tasks and optimizing the gap between them. The paradigm was firstly applied to correct the current ResNet protocol by simply reducing the stride and moving the Conv5 block from the head to the feature extraction network, which brings 3.6 improvement of AP70 on the PASCAL VOC dataset and 1.5 improvement of AP on the COCO dataset for ResNet50. Next, the new method is applied on the Faster R-CNN with backbone of VGG16,ResNet50 and ResNet101, which brings above 2.0 improvement of AP70 on the PASCAL VOC dataset and above 1.9 improvement of AP on the COCO dataset. Noticeably, the implementation of G-RCN only involves a few structural modifications, with no extra module added.