Abstract:With the rapid development of remote sensing technology and the growing demand for applications, the classical deep learning-based object detection model is bottlenecked in processing incremental data, especially in the increasing classes of detected objects. It requires models to sequentially learn new classes of objects based on the current model, while preserving old categories-related knowledge. Existing class-incremental detection methods achieve this goal mainly by constraining the optimization trajectory in the feature of output space. However, these works neglect the case where the previously learned background is a new category to learn, resulting in performance degradation in the new category because of the conflict between remaining the background-related knowledge or updating the background-related knowledge. This paper proposes a novel class-incremental detection method incorporated with the teacher-student architecture and the selective distillation (SDCID) strategy. Specifically, it is the asymmetry architecture, i.e., the teacher network temporarily stores historical knowledge of previously learned objects, and the student network integrates historical knowledge from the teacher network with the newly learned object-related knowledge, respectively. This asymmetry architecture reveals the significance of the distinct representation of history knowledge and new knowledge in incremental detection. Furthermore, SDCID selectively masks the shared subobject of new images to learn and previously learned background, while learning new categories of images and then transfers the classification results of the student model to the background class following the judgment model of the teacher model. This manner avoids interferences between the background category-related knowledge from a teacher model and the learning of other new classes of objects. In addition, we proposed a new incremental learning evaluation metric, C-SP, to comprehensively evaluate the incremental learning stability and plasticity performance. We verified the proposed method on two object detection datasets of remote sensing images, i.e., DIOR and DOTA. The experience results in accuracy and C-SP suggest that the proposed method surpasses existing class-incremental detection methods. We further analyzed the influence of the mask component in our method and the hyper-parameters sensitive to our method.

Incremental Deep Neural Network for Visual Categories

Efficient Incremental Training for Deep Convolutional Neural Networks

Incremental learning with neural networks for computer vision: a survey

Error-Driven Incremental Learning in Deep Convolutional Neural Network for Large-Scale Image Classification

Structurally Enhanced Incremental Neural Learning for Image Classification with Subgraph Extraction

Incremental Generalized Category Discovery

Class-Incremental Learning: Survey and Performance Evaluation on Image Classification

Hierarchical Gate Network for Fine-Grained Visual Recognition.

Tree-CNN: A Hierarchical Deep Convolutional Neural Network for Incremental Learning

Incremental Classification of Remote Sensing Images with Feature Pyramid and Class Hierarchy

ADINet: Attribute Driven Incremental Network for Retinal Image Classification

Progressive Neural Networks for Image Classification

Incremental Scene Classification Using Dual Knowledge Distillation and Classifier Discrepancy on Natural and Remote Sensing Images

Evolving Convolutional Neural Network And Its Application In Fine-Grained Visual Categorization

iNeMo: Incremental Neural Mesh Models for Robust Class-Incremental Learning

Class-Incremental Learning: A Survey

An Incremental Neural Network for Online Supervised Learning and Topology Learning

Incremental Learning In Online Scenario

A Class-Incremental Detection Method of Remote Sensing Images Based on Selective Distillation

Hierarchical Task-Incremental Learning with Feature-Space Initialization Inspired by Neural Collapse

Incremental Concept Formation over Visual Images Without Catastrophic Forgetting