2

This work introduces LymphAware, a domain-aware bias disruption framework designed to improve the robustness and clinical reliability of lymphoma histopathology AI systems. Modern classifiers often rely on shortcut signals such as scanner-specific color distributions, staining variability, and slide-preparation artifacts, which artificially inflate in-domain accuracy but collapse under cross-acquisition domain shift. LymphAware explicitly disentangles disease-relevant morphological features from acquisition-sensitive nuisance factors through three complementary mechanisms, morphology-centric feature isolation, adversarial and orthogonality-based shortcut suppression, and cross-domain stability regularization. To further expose hidden shortcut dependencies, the framework incorporates artifact-shift perturbations that simulate realistic staining and scanner variability while enforcing counterfactual consistency during training. Extensive evaluation on a heterogeneous multi-source lymphoma benchmark demonstrates improved cross-domain generalization, stable behavior under hyperparameter variation, and attribution maps better aligned with pathology-relevant regions. While these explanations reflect associative alignment rather than formal causal inference, the findings underscore the necessity of representation-level shortcut disruption for building clinically trustworthy lymphoma diagnostic AI systems.

Satellite image inpainting is a crucial task in remote sensing, where accurately restoring missing or occluded regions is essential for robust image analysis. In this paper, we propose KAO, a novel framework that utilizes Kernel-Adaptive Optimization within diffusion models for satellite image inpainting. KAO is specifically designed to address the challenges posed by very high-resolution (VHR) satellite datasets, such as DeepGlobe and the Massachusetts Roads Dataset. Unlike existing methods that rely on preconditioned models requiring extensive retraining or postconditioned models with significant computational overhead, KAO introduces a Latent Space Conditioning approach, optimizing a compact latent space to achieve efficient and accurate inpainting. Furthermore, we incorporate Explicit Propagation into the diffusion process, facilitating forward-backward fusion, which improves the stability and precision of the method. Experimental results demonstrate that KAO sets a new benchmark for VHR satellite image restoration, providing a scalable, high-performance solution that balances the efficiency of preconditioned models with the flexibility of postconditioned models.

GNSS data offers a reliable alternative for estimating Precipitable Water Vapor (PWV), but accurate GPS-PWV determination in tropical climates requires weighted mean temperature (Tm). With traditional measurement methods often unavailable in Thailand, and existing empirical models showing low accuracy, we propose a deep learning approach. Our Bidirectional Learning with Attention (BLA) model incorporates GRUs and an attention mechanism for Tm modeling. Trained on ERA5 data (2017-2021) and evaluated on 2022 data, BLA-Tm achieved 76% improvement over conventional models, reducing biases significantly. Validation with 280 GNSS stations confirmed BLA-Tm’s superior accuracy in GPS-PWV estimation.

In this paper, we present MeViT (Medium-Resolution Vision Transformer), designed for semantic segmentation of Landsat satellite imagery, focusing on key economic crops in Thailand para rubber, corn, and pineapple. MeViT enhances Vision Transformers (ViTs) by integrating medium-resolution multi-branch architectures and revising mixed-scale convolutional feedforward networks (MixCFN) to extract multi-scale local information. Extensive experiments on a public Thailand dataset demonstrate that MeViT outperforms state-of-the-art deep learning methods, achieving a precision of 92.22%, recall of 94.69%, F1 score of 93.44%, and mean IoU of 83.63%. These results highlight MeViT’s effectiveness in accurately segmenting Thai Landsat-8 data.

Detecting objects of varying sizes, like kilometer stones, remains a significant challenge and directly affects the accuracy of object counts. Transformers have shown remarkable success in natural language processing (NLP) and image processing due to their ability to model long-range dependencies. This paper proposes an enhanced YOLO (You Only Look Once) series with two key contributions, (i) We employ a pre-training objective to obtain original visual tokens from image patches of road assets, using a pre-trained Vision Transformer (ViT) backbone, which is then fine-tuned on downstream tasks with additional task layers. (ii) We incorporate Feature Pyramid Network (FPN) decoder designs into our deep learning network to learn the significance of different input features, avoiding issues like feature mismatch and performance degradation that arise from simple summation or concatenation. Our proposed method, Transformer-Based YOLOX with FPN, effectively learns general representations of objects and significantly outperforms state-of-the-art detectors, including YOLOv5S, YOLOv5M, and YOLOv5L. It achieves a 61.5% AP on the Thailand highway corpus, surpassing the current best practice (YOLOv5L) by 2.56% AP on the test-dev dataset.

Flooding poses a significant challenge in Thailand due to its complex geography, traditionally addressed through GIS methods like the Flood Risk Assessment Model (FRAM) combined with the Analytical Hierarchy Process (AHP). This study assesses the efficacy of Artificial Neural Networks (ANN) in flood susceptibility mapping, using data from Ayutthaya Province and incorporating 5-fold cross-validation and Stochastic Gradient Descent (SGD) for training. ANN achieved superior performance with precision of 79.90%, recall of 79.04%, F1-score of 79.08%, and accuracy of 79.31%, outperforming the traditional FRAM approach. Notably, ANN identified that only three factors—flow accumulation, elevation, and soil types—were crucial for predicting flood-prone areas. This highlights the potential for ANN to simplify and enhance flood risk assessments. Moreover, the integration of advanced machine learning techniques underscores the evolving capability of AI in addressing complex environmental challenges.