Search

In this paper, we present a novel dimensionality reduction term (i.e. TripletPCA and ContrastivePCA) designed to enhance the robustness of pair-based loss functions in metric learning against image corruptions. Notably, our approach achieves this without the need for any data preprocessing steps. Our method can be seamlessly integrated into existing models, such as a block with Plug & Play framework. To the best of our knowledge, our TripletPCA and ContrastivePCA represent the first attempts to incorporate dimensionality reduction directly into pair-based metric learning losses for end-to-end metric learning. By projecting image features into low-dimensional vectors, our proposed loss functions effectively retain the essential components of images while mitigating the impact of corrupted features. Consequently, our metric learning loss function accurately computes feature distances through the projection. Experimental results demonstrate that our dimensionality reduction term can be easily incorporated into various types of existing deep neural networks. This integration leads to a substantial improvement in performance on standard benchmark datasets for corrupted image classification tasks. On several corruption datasets, we achieve an average performance improvement of 10.55% compared to existing baseline methods. Our code is available at https://github.com/Juryun/TDRL. [ABSTRACT FROM AUTHOR]

Automatic classifier accuracy evaluation (ACAEval) on unlabeled test sets is critical for unseen real-world environments. The use of dataset-level regression on synthesized meta-datasets (comprised of many sample sets) has shown promising results for ACAEval. However, the existing meta-dataset for ACAEval is created using simple image transformations such as rotation and background substitution, which can make it difficult to ensure a reasonable distribution shift between the sample set and the test set. When the distribution shift is large, it becomes challenging to estimate the classifier accuracy on the test set using those sample sets. To ensure more robust ACAEval, this paper attempts to customize a meta-dataset in which each sample set has a reasonable distribution shift to the test set. An intra-class cycle-consistent adversarial learning (ICAL) method is introduced to transfer the style of a labeled training set to the style of the test set, by jointly considering the domain shift issue, the label flipping issue (the semantic information may be changed after style transformation), and the diversity of multiple sample sets in the meta-dataset. Experiments validate that under the same experimental setup, our method outperforms the existing ACAEval methods by a good margin, and achieves state-of-the-art performance on several standard benchmark datasets, including digit classification and natural image classification. • Two authors named Yan Huang in pinyin, one PostDoc (1st author), the other an Associate Prof (3rd author). • Paper on ACAEval for classifier accuracy. Uses meta-dataset technique for unlabeled real-world data. • Customized meta-dataset for ACAEval to address distribution shift between samples and test set. • Our sample set considers label flip issue. Introduces random indicator and FD margin loss. • Experiments demonstrate our ACAEval matches existing methods and surpasses dataset-level regression. [ABSTRACT FROM AUTHOR]

• To apply ViT to adaptively focus on more discriminative regions in aerial images, a novel transformer-based method, called D-BiT, is proposed, which achieves SOTA performance on several standard benchmarks. • To address the fitting dilemma of ViT caused by the lack of trainable aerial images, a Discriminative Label Smoothing Module (DLSM) is designed to enhance ViT to fully learn discriminative features without overfitting the distribution of noisy data, even in the condition of insufficient data volume. • Aiming at aerial images with complex backgrounds, a discriminative supervised contrastive loss is proposed to effectively combine label information and achieve a more compact and reasonable intra-class structure. Aerial image recognition (AIR) has received increasing interest in the remote sensing community, and complex backgrounds with various types of objects bring huge challenges to recognition. Vision transformer (ViT) can capture long-range contextual relations, but it tends to fall into over-fitting for training samples. This problem is particularly acute in AIR where trainable data is insufficient. While training ViT with label smoothing can alleviate this problem, it is easy to pose the risk of underfitting and relies heavily on data quality. Furthermore, the application of ViT in AIR has always been limited because aerial images generally have greater intra-class variation and inter-class similarity than natural images. Thus, we propose a ViT-based framework called D-BiT to extend ViT to learn discriminative features in aerial images by integrating two heads on the ViT backbone. One is a classifier head with the Discriminative Label Smoothing Module (DLSM) to fully learn effective features without overfitting the distribution of noisy data, even in the condition of insufficient data volume. Another is a projector head with a newly-designed supervised contrastive loss to effectively combine label information and achieve a more compact and reasonable intra-class structure. Experimental results on three popular AIR benchmarks demonstrate the effectiveness of D-BiT: Aerial Image Dataset (AID), UC Merced and Event Recognition in Aerial videos (ERA), where we achieve state-of-the-art (SOTA) performance. In particular, D-BiT can significantly improve the accuracy of ViT for recognizing aerial images. [ABSTRACT FROM AUTHOR]

Existing scene text recognition methods have incorporated text rectification to lessen text irregularity in images for accurate recognition. Previous text rectification methods aim to convert an irregular text image into a regular form, making it easier to be recognized. In this study, we explore text rectification for text recognition and discover the issues: performance degradation of the recognition network and the unreliable situation of text rectification, which are ignored by all previous works. Therefore, we rethink what is causing two issues, and propose a rectification-based text recognition network to mitigate the above issues. The proposed network consists of text rectification and text recognition, and designs a multi-level feature aggregation module to enhance feature learning for character representation. Concretely, we devise a mixed batch training strategy to address the performance degradation of the recognition network, and design a confidence decoding scheme to avoid the unreliable situation of text rectification. Extensive ablation studies verified the positive role of the feature aggregation module in feature learning and the effectiveness of the proposed training strategy and decoding scheme in addressing the issues. Experimental results outperform the state-of-the-art results on public benchmarks. • We explore the limitation and unreliable cases caused by the rectification network. • Mixed batch and confidence decoding are designed for accurate recognition. • We design feature aggregation to exploit multi-level features. • Our method achieves superior performance on standard datasets. [ABSTRACT FROM AUTHOR]