Decoupled Self Attention for Accurate One Stage Object Detection
This work offers an incremental improvement in accuracy for one-stage object detection models, specifically benefiting researchers and practitioners working with RetinaNet and similar architectures.
The paper addresses the deviation between features extracted by ImageNet-pretrained backbones and the features required for object localization in one-stage detectors. They propose a Decoupled Self Attention (DSA) module that extracts task-appropriate features, leading to performance increases of 0.4% AP (ResNet50) and 0.5% AP (ResNet101) on the COCO dataset for RetinaNet. When combined with an object confidence task, these gains further increase to 1.0% AP and 1.4% AP respectively.
As the scale of object detection dataset is smaller than that of image recognition dataset ImageNet, transfer learning has become a basic training method for deep learning object detection models, which will pretrain the backbone network of object detection model on ImageNet dataset to extract features for classification and localization subtasks. However, the classification task focuses on the salient region features of object, while the location task focuses on the edge features of object, so there is certain deviation between the features extracted by pretrained backbone network and the features used for localization task. In order to solve this problem, a decoupled self attention(DSA) module is proposed for one stage object detection models in this paper. DSA includes two decoupled self-attention branches, so it can extract appropriate features for different tasks. It is located between FPN and head networks of subtasks, so it is used to extract global features based on FPN fused features for different tasks independently. Although the network of DSA module is simple, but it can effectively improve the performance of object detection, also it can be easily embedded in many detection models. Our experiments are based on the representative one-stage detection model RetinaNet. In COCO dataset, when ResNet50 and ResNet101 are used as backbone networks, the detection performances can be increased by 0.4% AP and 0.5% AP respectively. When DSA module and object confidence task are applied in RetinaNet together, the detection performances based on ResNet50 and ResNet101 can be increased by 1.0% AP and 1.4% AP respectively. The experiment results show the effectiveness of DSA module. Code is at: https://github.com/chenzuge1/DSANet.git.