CILEA-NET: Curriculum-Based Incremental Learning Framework for Remote Sensing Image Classification

Abstract

In this paper, we address class incremental learning (IL) in remote sensing image analysis. Since remote sensing images are acquired continuously over time by Earth's Observation sensors, the land-cover/land-use classes on the ground are likely to be found in a gradational manner. This process restricts the deployment of stand-alone classification approaches, which are trained for all the classes together in one iteration. Therefore, for every new set of categories discovered, the entire network consisting of old and new classes requires retraining. This procedure is often impractical, considering vast volumes of data, limited resources, and the complexity of learning models. In this respect, we propose a ConvNet based framework (called CILEA-NET, Curriculum-based Incremental LEArning Framework for Remote Sensing Image Classification) to efficiently resolve the difficulties associated with incremental learning paradigm. The framework includes new classes in the already trained model to avoid catastrophic forgetting for the old while ensuring improved generalization for the newly added classes. To manage the IL's stability-plasticity dilemma, we introduce a novel curriculum learning-based approach where the order of the new classes is devised based on their similarity to the already trained classes. We then perform the training in that given order. We notice that the curriculum learning setup distinctly enhances the training time for the new classes. Experimental results on several optical datasets: PatternNet and NWPU-RESISC45, and a hyperspectral dataset: Indian Pines, validate the robustness of our technique.

Anthropogenic Object Localization: Evaluation of Broad-Area High-Resolution Imagery Scans Using Deep Learning in Overhead Imagery

“…In recent years, many benchmark datasets have been released, and these datasets have been researched with a high level of success, such as in [ 8 , 19 , 20 , 21 ]. With these datasets, a high level of average recall can be obtained through the careful selection of hyperparameters and augmentation schemes.…”

Continual Barlow Twins: Continual Self-Supervised Learning for Remote Sensing Semantic Segmentation

“…Continual learning benchmark for remote sensing [64] is a large-scale remote sensing image scene classification database based on three CL scenarios. CILEA-Net [65] proposes a CL strategy, based on the incremental learning of new classes ordered according to the similarity with the old ones. In [66], an incremental learning with open-set recognition framework and a new loss are proposed for RS image scene classification.…”

Color Image Quality Assessment with Multi Deep Convolutional Networks

“…Incremental learning is a technique that allows a neural network to continuously update its parameters with incremental data, breaking the traditional one-off training process in deep learning. IL has been explored in various fields, including computer vision [7], [8], natural language processing [9], [10] and remote sensing [11].The most significant challenge in IL is catastrophic forgetting, which occurs when the parameter updates result in the loss of previously learned knowledge. This phenomenon was first identified and discussed as early as the 1980s by McCloskey, et al [12].…”

“…a : {[0, 1,2,3,4,5,6,7,8,9,10,11,12,13,14,15], [16], [17], [18], [19], [20]} b : {[0, 12,9,20,7,15,8,14,16,5,19,4,1,13,2,11], [17], [3], [6], [18], [10]} c : {[0, 13,19,15,17,9,8,5,20,4,3,10,11,18,…”