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 Active Learning Method For Supervised Isomap

Incremental semi-supervised subspace learning for image retrieval.

A Novel Approach to Incremental SVM Learning Algorithm

An Incremental Updating Method for Support Vector Machines

Incremental learning algorithm based on support vector machine

Supervised nonlinear dimensionality reduction for visualization and classification.

New Incremental Learning Algorithm for Semi-Supervised Support Vector Machine

Semi-supervised Discriminant Isomap with Application to Visualization, Image Retrieval and Classification

Incremental Semantic Mapping with Unsupervised On-line Learning

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

Tackle Balancing Constraint for Incremental Semi-Supervised Support Vector Learning

Incremental Learning With Open-Set Recognition for Remote Sensing Image Scene Classification

Active Learning in the Spatial Domain for Remote Sensing Image Classification

Active Learning for Improved Semi-Supervised Semantic Segmentation in Satellite Images

IGSOM: Incremental Clustering Based on Self-Organizing-Mapping

An incremental online semi-supervised active learning algorithm based on self-organizing incremental neural network

Incremental Self-training for Semi-supervised Learning

Incremental learning with online SVMs on LiDAR sensory data

Incremental learning algorithm for large-scale semi-supervised ordinal regression

A survey of active learning algorithms for supervised remote sensing image classification

An Incremental SVM Learning Algorithm Α-Isvm