TAE-Net: Task-Adaptive Embedding Network for Few-Shot Remote Sensing Scene Classification

Abstract

Recently, approaches based on deep learning are quite prevalent in the area of remote sensing scene classification. Though significant success has been achieved, these approaches are still subject to an excess of parameters and extremely dependent on a large quantity of labeled data. In this study, few-shot learning is used for remote sensing scene classification tasks. The goal of few-shot learning is to recognize unseen scene categories given extremely limited labeled samples. For this purpose, a novel task-adaptive embedding network is proposed to facilitate few-shot scene classification of remote sensing images, referred to as TAE-Net. A feature encoder is first trained on the base set to learn embedding features of input images in the pre-training phase. Then in the meta-training phase, a new task-adaptive attention module is designed to yield the task-specific attention, which can adaptively select informative embedding features among the whole task. In the end, in the meta-testing phase, the query image derived from the novel set is predicted by the meta-trained model with limited support images. Extensive experiments are carried out on three public remote sensing scene datasets: UC Merced, WHU-RS19, and NWPU-RESISC45. The experimental results illustrate that our proposed TAE-Net achieves new state-of-the-art performance for few-shot remote sensing scene classification.

Collaborative Self-Supervised Transductive Few-Shot Learning for Remote Sensing Scene Classification

“…Additionally, the combination of information from other samples in the dataset and the use of transductive inference contribute to its performance. [26] 85.41 ± 0.35 92.28 ± 0.13 baseline [33] 75.57 ± 0.36 88.65 ± 0.18 S2M2 [55] 69.00 ± 0.41 82.14 ± 0.21 Meta-SGD [56] 51.54 ± 2.31 61.74 ± 2.02 MatchingNet [57] 76.14 ± 0.35 84.00 ± 0.20 ProtoNet [58] 77.00 ± 0.36 91.70 ± 0.15 RelationNet [59] 77.76 ± 0.34 86.84 ± 0.15 DLA-MatchNet [35] 70.21 ± 0.32 81.86 ± 0.52 IDLN [36] 73.89 ± 0.88 83.12 ± 0.56 TAE-Net [37] 73.67 ± 0.74 88.95 ± 0.52 CAN+T [60] 69.79 ± 0.56 79.71 ± 0.22 SPNet [25] 81.06 ± 0.60 88.04 ± 0.28 DANet [28] 75.02 ± 0.16 89.21 ± 0.07 CS 2 TFSL (Ours) 86.03 ± 0.13 93.09 ± 0.05…”

“…It consists of two main components: an embedding network and an iterative distribution learning strategy. Huang et al [37] proposed a task-adaptive attention component, which combined the meta-learning training mechanism and graph neural network transductive inference in few-shot scene classification.…”

“…We selected the SGD (Stochastic Gradient Descent) optimizer with a momentum of 0.9. In terms of comparative algorithms, we selected 14 classical and stateof-the-art few-shot classification algorithms, namely MAML [54], LLSR [27], Ji et al [26], baseline [33], S2M2 [55], Meta-SGD [56], MatchingNet [57], ProtoNet [58], RelationNet [59], DLA-MatchNet [35], TAE-Net [37], IDLN [36], CAN+T [60], SPNet [25], and DANet [28]. For fair comparison, all comparison methods were reported from their respective papers and taken from the final experimental results.…”

Improved prototypical networks for remote sensing scene classification

“…The experimental results on the Single Datasets are listed in Table 2. The proposed method is compared with RDPN [17], TAE-Net [18], SAFFNet, DLA MatchNet, SCL-MLNet [19], L2-Norm ProtoNet, and RelationNet [20]. It can be seen from Table 2 that the accuracy of our method on the three RS scene datasets is significantly higher than all other methods.…”

Fuzzy Cognitive Maps with Bird Swarm Intelligence Optimization-Based Remote Sensing Image Classification

“…Huang and others [ 15 ] presented a task-adoptive embedding network for facilitating few-shot scene classification of RSI, represented as TAE-Net. First, a feature encoder was trained on the base set for learning embedded features of input image in the pretraining stage.…”