Benchmarking the Alignment of Data-Quality Metrics, Human Judgment and Land-Cover Segmentation Performance for Earth Observation

Volume and quality of datasets are crucial for deep learning model training, yet they are often constrained by availability and data acquisition costs. Synthetic data augmentation can extend existing datasets with realistic images, and the quality of these images is generally assessed through fidelity metrics such as FID, KID, IS, LPIPS and SSIM that measure structural or distributional similarity. However, such metrics, including the widely used FID, focus on visual fidelity without reflecting downstream utility, and can diverge from human perception under perturbations that are imperceptible to human observers. In this work, we systematically evaluate Earth observation datasets alongside synthetic counterparts generated by deep generative models, comparing automatic metrics against human perception and downstream tasks. Our results reveal a stark misalignment: semantics-preserving perturbations such as rotation drastically alter metric scores while leaving human recognition unaffected, and synthetic samples that score poorly on automatic metrics achieve comparable or higher perceived realism, and can improve downstream performance when combined with real data. By benchmarking semantic segmentation models trained on mixed real-synthetic datasets, we demonstrate that quality metrics rooted in ImageNet-pretrained feature spaces are unreliable indicators for geospatial data. Our findings underscore that automatic quality evaluation of synthetic datasets should be grounded in downstream task performance and human evaluation.

BELDE: Building a Large-scale Earth-observation Land-cover Dataset for Europe

Earth observation imagery plays a critical role in environmental monitoring, urban planning, disaster assessment, and climate analysis. While multi-spectral sensors are increasingly available, true-color (RGB) imagery remains widely used due to the power, cost, and deployment constraints of many satellite and aerial platforms. However, existing land-cover segmentation datasets are often limited in geographic coverage, scale, or public accessibility. To bridge this gap, we introduce BELDE (Building a Large-scale Earth-observation Land-cover Dataset for Europe), a publicly available dataset tailored for RGB-based remote sensing semantic segmentation. Constructed from Sentinel-2 true-color images and ESA WorldCover data annotations, BELDE contains 1,088,385 curated image-segmentation map pairs spanning Europe with 7 land-cover classes at 10 m spatial resolution, making it one of the largest publicly available RGB land-cover segmentation datasets for Earth observation. To facilitate cross-region generalization studies, we additionally introduce BELDE-K (16,607 pairs) covering the Republic of Korea and BELDE-CA-NV (88,155 pairs) covering California and Nevada in the United States. We establish baseline results using multiple semantic segmentation architectures and evaluate both in-domain and cross-domain performance. Models trained on BELDE achieve an F1 score of 83.0% on the European test set, while performance decreases to 66.4% on BELDE-CA-NV and 58.3% on BELDE-K, highlighting the challenges posed by out-of-distribution geographic domain shift. By providing a continental-scale RGB segmentation and evaluation benchmark, BELDE supports the development of robust and transferable Earth observation models. The dataset and benchmark resources will be publicly released.

LALE: Lightweight-Transformer Architecture for Land-Cover Estimation

Semantic segmentation of remote sensing imagery requires models that capture both global context and local detail under tight computational budgets. Prior work typically optimizes for one of these axes: attention for global context, convolution for local detail, or compactness for efficiency. While hybrid approaches aim to capture both, they require architectural changes and encoder backbones with computational overhead, limiting efficiency and performance. We present LALE (Lightweight-transformer Architecture for Land-cover Estimation), an end-to-end remote sensing image segmentation architecture, that bifurcates its encoder by resolution: lightweight ConvMixer stages handle high-resolution local features, while transformer stages handle low-resolution global context, confining the quadratic cost of self-attention to deep, downsampled feature maps. An all-MLP multi-scale decoder, together with RMSNorm and StarReLU throughout, further reduces compute and parameter count. On the large-scale ARAS400k remote-sensing segmentation benchmark, LALE establishes a strong efficiency-performance trade-off against CNN, transformer, and hybrid baselines. Our smallest variant, (just 1.6M parameters), reaches within 2.6 F1 points of the best baseline (UPerNet) while using 4.5x fewer parameters, 7x less storage, 17x fewer GMACs, and delivering 1.8x higher throughput.

Disentangled Anatomy-Disease Diffusion (DADD) for Controllable Ulcerative Colitis Progression Synthesis

Synthesizing longitudinal medical images at controllable disease stages while preserving patient-specific anatomy is hindered by the entanglement of pathological textures and structural features. We address this challenge for ulcerative colitis (UC) endoscopy, where severity follows a continuous ordinal progression along the Mayo Endoscopic Score (MES). Our framework, Disentangled Anatomy-Disease Diffusion (DADD), conditions a latent diffusion model on two complementary embeddings: a pretrained image encoder for patient anatomy and a separately trained ordinal embedder for cumulative disease severity. Since image embeddings inevitably capture disease information, we introduce a Feature Purifier, a cross-attention-based erasure mechanism that identifies and suppresses disease-correlated channels, yielding purified anatomical representations. These cleaned anatomy tokens and target disease tokens are injected into the denoising network via a Triple-Pathway Cross-Attention mechanism with resolution-dependent routing gates. This architecture leverages the U-Net hierarchy, in which different network depths encode global structure versus fine-grained pathological texture. Furthermore, we introduce Delta Steering, a training-free directional signal derived from the ordinal embeddings that enables explicit, single-pass control over disease transitions at inference without requiring additional forward passes. Validated on the LIMUC dataset, our approach produces high-fidelity images across all severity levels and effectively rebalances skewed class distributions, enhancing performance for downstream classification tasks. The dataset is available at zenodo.org/records/5827695 and the code base at github.com/umutdundar99/progressive-stable-diffusion

SenBen: Sensitive Scene Graphs for Explainable Content Moderation

Content moderation systems classify images as safe or unsafe but lack spatial grounding and interpretability: they cannot explain what sensitive behavior was detected, who is involved, or where it occurs. We introduce the Sensitive Benchmark (SenBen), the first large-scale scene graph benchmark for sensitive content, comprising 13,999 frames from 157 movies annotated with Visual Genome-style scene graphs (25 object classes, 28 attributes including affective states such as pain, fear, aggression, and distress, 14 predicates) and 16 sensitivity tags across 5 categories. We distill a frontier VLM into a compact 241M student model using a multi-task recipe that addresses vocabulary imbalance in autoregressive scene graph generation through suffix-based object identity, Vocabulary-Aware Recall (VAR) Loss, and a decoupled Query2Label tag head with asymmetric loss, yielding a +6.4 percentage point improvement in SenBen Recall over standard cross-entropy training. On grounded scene graph metrics, our student model outperforms all evaluated VLMs except Gemini models and all commercial safety APIs, while achieving the highest object detection and captioning scores across all models, at $7.6\times$ faster inference and $16\times$ less GPU memory.

D-GATNet: Interpretable Temporal Graph Attention Learning for ADHD Identification Using Dynamic Functional Connectivity

Attention Deficit Hyperactivity Disorder (ADHD) is a prevalent neurodevelopmental disorder whose neuroimaging-based diagnosis remains challenging due to complex time-varying disruptions in brain connectivity. Functional MRI (fMRI) provides a powerful non-invasive modality for identifying functional alterations. Existing deep learning (DL) studies employ diverse neuroimaging features; however, static functional connectivity remains widely used, whereas dynamic connectivity modeling is comparatively underexplored. Moreover, many DL models lack interpretability. In this work, we propose D-GATNet, an interpretable temporal graph-based framework for automated ADHD classification using dynamic functional connectivity (dFC). Sliding-window Pearson correlation constructs sequences of functional brain graphs with regions of interest as nodes and connectivity strengths as edges. Spatial dependencies are learned via a multi-layer Graph Attention Network, while temporal dynamics are modeled using 1D convolution followed by temporal attention. Interpretability is achieved through graph attention weights revealing dominant ROI interactions, ROI importance scores identifying influential regions, and temporal attention emphasizing informative connectivity segments. Experiments on the Peking University site of the ADHD-200 dataset using stratified 10-fold cross-validation with a 5-seed ensemble achieved 85.18% +_5.64 balanced accuracy and 0.881 AUC, outperforming state-of-the-art methods. Attention analysis reveals cerebellar and default mode network disruptions, indicating potential neuroimaging biomarkers.

DuSCN-FusionNet: An Interpretable Dual-Channel Structural Covariance Fusion Framework for ADHD Classification Using Structural MRI

Attention Deficit Hyperactivity Disorder (ADHD) is a highly prevalent neurodevelopmental condition; however, its neurobiological diagnosis remains challenging due to the lack of reliable imaging-based biomarkers, particularly anatomical markers. Structural MRI (sMRI) provides a non-invasive modality for investigating brain alterations associated with ADHD; nevertheless, most deep learning approaches function as black-box systems, limiting clinical trust and interpretability. In this work, we propose DuSCN-FusionNet, an interpretable sMRI-based framework for ADHD classification that leverages dual-channel Structural Covariance Networks (SCNs) to capture inter-regional morphological relationships. ROI-wise mean intensity and intra-regional variability descriptors are used to construct intensity-based and heterogeneity-based SCNs, which are processed through an SCN-CNN encoder. In parallel, auxiliary ROI-wise variability features and global statistical descriptors are integrated via late-stage fusion to enhance performance. The model is evaluated using stratified 10-fold cross-validation with a 5-seed ensemble strategy, achieving a mean balanced accuracy of 80.59% and an AUC of 0.778 on the Peking University site of the ADHD-200 dataset. DuSCN-FusionNet further achieves precision, recall, and F1-scores of 81.66%, 80.59%, and 80.27%, respectively. Moreover, Grad-CAM is adapted to the SCN domain to derive ROI-level importance scores, enabling the identification of structurally relevant brain regions as potential biomarkers.

Grounding Synthetic Data Generation With Vision and Language Models

Deep learning models benefit from increasing data diversity and volume, motivating synthetic data augmentation to improve existing datasets. However, existing evaluation metrics for synthetic data typically calculate latent feature similarity, which is difficult to interpret and does not always correlate with the contribution to downstream tasks. We propose a vision-language grounded framework for interpretable synthetic data augmentation and evaluation in remote sensing. Our approach combines generative models, semantic segmentation and image captioning with vision and language models. Based on this framework, we introduce ARAS400k: A large-scale Remote sensing dataset Augmented with Synthetic data for segmentation and captioning, containing 100k real images and 300k synthetic images, each paired with segmentation maps and descriptions. ARAS400k enables the automated evaluation of synthetic data by analyzing semantic composition, minimizing caption redundancy, and verifying cross-modal consistency between visual structures and language descriptions. Experimental results indicate that while models trained exclusively on synthetic data reach competitive performance levels, those trained with augmented data (a combination of real and synthetic images) consistently outperform real-data baselines. Consequently, this work establishes a scalable benchmark for remote sensing tasks, specifically in semantic segmentation and image captioning. The dataset is available at zenodo.org/records/18890661 and the code base at github.com/caglarmert/ARAS400k.

Local-Global Prompt Learning via Sparse Optimal Transport

Few-shot adaptation of vision-language models (VLMs) like CLIP typically relies on learning textual prompts matched to global image embeddings. Recent works extend this paradigm by incorporating local image-text alignment to capture fine-grained visual cues, yet these approaches often select local regions independently for each prompt, leading to redundant local feature usage and prompt overlap. We propose SOT-GLP, which introduces a shared sparse patch support and balanced optimal transport allocation to explicitly partition salient visual regions among class-specific local prompts while preserving global alignment. Our method learns shared global prompts and class-specific local prompts. The global branch maintains standard image-text matching for robust category-level alignment. The local branch constructs a class-conditioned sparse patch set using V-V attention and aligns it to multiple class-specific prompts via balanced entropic optimal transport, yielding a soft partition of patches that prevents prompt overlap and collapse. We evaluate our method on two complementary objectives: (i) few-shot classification accuracy on 11 standard benchmarks and (ii) out-of-distribution (OOD) detection. On the standard 11-dataset benchmark with 16-shot ViT-B/16, SOT-GLP achieves 85.1% average accuracy, outperforming prior prompt-learning methods. We identify a distinct accuracy-robustness trade-off in prompt learning: while learnable projections optimize in-distribution fit, they alter the foundational feature space. We demonstrate that a projection-free local alignment preserves the native geometry of the CLIP manifold, yielding state-of-the-art OOD detection performance (94.2% AUC) that surpasses fully adapted models. Implementation available at: https://github.com/Deniz2304988/SOT-GLP

TopoMaskV3: 3D Mask Head with Dense Offset and Height Predictions for Road Topology Understanding

Mask-based paradigms for road topology understanding, such as TopoMaskV2, offer a complementary alternative to query-based methods by generating centerlines via a dense rasterized intermediate representation. However, prior work was limited to 2D predictions and suffered from severe discretization artifacts, necessitating fusion with parametric heads. We introduce TopoMaskV3, which advances this pipeline into a robust, standalone 3D predictor via two novel dense prediction heads: a dense offset field for sub-grid discretization correction within the existing BEV resolution, and a dense height map for direct 3D estimation. Beyond the architecture, we are the first to address geographic data leakage in road topology evaluation by introducing (1) geographically distinct splits to prevent memorization and ensure fair generalization, and (2) a long-range (+/-100 m) benchmark. TopoMaskV3 achieves state-of-the-art 28.5 OLS on this geographically disjoint benchmark, surpassing all prior methods. Our analysis shows that the mask representation is more robust to geographic overfitting than Bezier, while LiDAR fusion is most beneficial at long range and exhibits larger relative gains on the overlapping original split, suggesting overlap-induced memorization effects.