Abstract
As the object detection dataset scale is smaller than the image recognition dataset ImageNet scale, transfer learning has become a basic training method for deep learning object detection models, which pre-trains the backbone network of the object detection model on an ImageNet dataset to extract features for detection tasks. However, the classification task of detection focuses on the salient region features of an object, while the location task of detection focuses on the edge features, so there is a certain deviation between the features extracted by a pretrained backbone network and those needed by a localization task. To solve this problem, a decoupled self-attention (DSA) module is proposed for one-stage object-detection models in this paper. A DSA includes two decoupled self-attention branches, so it can extract appropriate features for different tasks. It is located between the Feature Pyramid Networks (FPN) and head networks of subtasks, and used to independently extract global features for different tasks based on FPN-fused features. Although the DSA network module is simple, it can effectively improve the performance of object detection, and can easily be embedded in many detection models. Our experiments are based on the representative one-stage detection model RetinaNet. In the Common Objects in Context (COCO) dataset, when ResNet50 and ResNet101 are used as backbone networks, the detection performances can be increased by 0.4 and 0.5% AP, respectively. When the DSA module and object confidence task are both applied in RetinaNet, the detection performances based on ResNet50 and ResNet101 can be increased by 1.0 and 1.4% AP, respectively. The experiment results show the effectiveness of the DSA module.
Highlights
IntroductionThe ImageNet [1] dataset is a large-scale image-classification dataset built by Professor
The ImageNet [1] dataset is a large-scale image-classification dataset built by ProfessorLi Fei-Fei, which contains tens of millions of images and tens of thousands of categories of objects
As the classification and localization subtasks in the object detection model focus on the different spatial features of objects, we propose a decoupled self-attention (DSA) module based on a self-attention mechanism to extract different spatial attention features for two subtasks
Summary
The ImageNet [1] dataset is a large-scale image-classification dataset built by Professor. In order to use transfer learning in object detection, we propose a method to extract features from pretrained features by introducing a self-attention mechanism, which can automatically extract relevant features for a specific task. The smaller receptive field of convolution features resulted in the pixel from the same object having different classes, which restricts the segmentation performance It used spatial and channel self-attention modules to extract features containing global spatial and channel information, which effectively improved segmentation performance. We propose an object detection model, DSANet, based on the self-attention mechanism, and uses the spatial domain self-attention module decoupled self-attention (DSA) to extract suitable features for specific tasks. FPN waspathway greatly weakened by order to improve the detection performance significantly increasing model padozens or hundreds of convolution layers,without so it proposed a bottom-up pathway for rameters and computation, a multiscale object detection method, Scale-Transferrable. In order to reduce human experience in head network design, the head network in this paper uses a self-attention mechanism to automatically extract features for each task
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Disclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.