Remote sensing scene classification (RSSC) is traditionally formulated as a vision-centric problem, where inference relies on rich visual representations extracted from images. In this work, we revisit this assumption by introducing text-only RSSC, a new problem setting that investigates whether remote sensing scenes can be recognized using only textual semantic representations. To enable this study, we propose SeProRS, which reformulates RSSC as a text-only classification task through semantic proxy generation. Specifically, a vision-language model (VLM) is first employed to convert remote sensing images into textual descriptions, termed semantic proxies, which are then encoded by a text embedding model for scene classification. This decoupled pipeline provides a unified and reproducible benchmark for evaluating text-only scene understanding. Furthermore, we investigate representative prompting strategies and proxy generators, and analyze their impact on classification performance, text embedding behavior, cross-generator transferability, and efficiency. Extensive experiments on three well-known RSSC datasets show that semantic proxies can preserve substantial discriminative information, while also revealing inherent limitations of purely text-based scene understanding. This paper does not aim to replace image-based RSSC, but instead establishes a new research direction by quantifying the feasibility and boundaries of semantic proxy-based scene recognition. The proposed benchmark and analysis provide a foundation for future research on efficient and modality-reduced remote sensing scene understanding. The source code and datasets will be accessible at https://github.com/sstary/SSRS.
Geospatial foundation models are pretrained on satellite and aerial imagery, but whether such an encoder beats a strong generic one for few-label land-use/land-cover mapping remains unclear, as prior comparisons rarely isolate the pretraining data. We study five matched pairs of Earth-observation (EO) and generic encoders sharing objective and architecture (DINO/SatDINO, DINOv3/DINOv3-SAT, MAE/SatMAE, MAE/Scale-MAE, CLIP/RemoteCLIP) within a benchmark of 14 frozen encoders on five RGB scene-classification datasets (0.06-30 m GSD). With native preprocessing, we evaluate k-NN, linear and attentive probes and image retrieval over six low-label splits, using paired per-image significance tests. EO pretraining is not uniformly helpful: on most datasets and metrics, it significantly improves three pairs (SatDINO, Scale-MAE, RemoteCLIP, the latter evaluated only on unseen datasets) and significantly hurts two (linear probe: DINOv3-SAT by up to 22.5 points, SatMAE by up to 8.8), with opposite outcomes even for two MAEs pretrained on the same fMoW-RGB imagery. These effects are robust to patch-token pooling and attentive probing, concentrate on small man-made objects, and are largest with 1-5% labels. Across encoders, the silhouette score of frozen embeddings predicts k-NN accuracy and retrieval mAP almost perfectly (Spearman $ρ\geq 0.96$ on four of five datasets). Pretraining on satellite imagery is thus not a reliable selection criterion; candidate checkpoints must be validated on target data.
Flow-matching models can produce high-quality image restorations, but their learned dynamics may amplify small input perturbations and make reconstructions unstable. We introduce LipsFlower to reduce this amplification. We derive stability bounds for Flow-from-$Y$, which initializes the ordinary differential equation from a noise-perturbed observation, as well as for observation-conditioned posterior sampling, which conditions the velocity field on the observation throughout integration. Based on this analysis, we develop LipHDiT, a Lipschitz-aware hourglass diffusion transformer that serves as the vector-field architecture of LipsFlower. LipHDiT incorporates stabilized approximate spectral normalization, bounded modulation, and convex residual mixing. We also use adversarial finite-radius gain regularization at sampled interpolation states to penalize excessive sensitivity to the state and, for conditional sampling, the observation. We test both formulations on blind face restoration, measuring adversarial robustness and the perceptual quality of the reconstructions. LipsFlower reduces perturbation amplification while keeping restoration quality competitive with the evaluated baselines. These results suggest that Lipschitz-aware flow matching can improve the stability of generative image restoration.
Remote photoplethysmography (rPPG) has been proposed as a physiological cue for deepfake detection, based on the premise that synthesized faces lack a genuine cardiac signal. We examine what rPPG-style detectors learn on Celeb-DF++ (22 forgery methods) and FaceForensics++ (FF++), using one identically sampled window per real and fake video. On talking-face forgeries, simple chrominance traces reach an AUC of 0.998, exceeding a dedicated rPPG extractor (0.87). Controlled experiments indicate that this signal is not physiological in origin. On fully synthesized forgeries, it is equally discriminative when extracted from background regions or from full frames without face detection, and it is concentrated above the 0.7 to 4 Hz heart-rate band. It is largely removed by shuffling frame order but persists after normalizing resolution, frame rate and encoding, and an independent reimplementation reproduces these findings. Removing high-frequency content makes most synthesized videos appear real, adding it makes real videos appear fake, and injecting a synthetic pulse has little effect. In controlled self-reenactment experiments with two generators, animating a face from a single frame introduces the same high-frequency signal, whereas reconstructing each frame from its own source does not. Face-based chrominance detectors trained on Celeb-DF++ classify over 99% of real FF++ videos as fake, indicating that they largely encode video source rather than manipulation, and a CLIP-based detector exhibits the same behavior. A frozen CLIP backbone matches or exceeds fusions of CLIP with rPPG and frequency features. Finally, we show that the official Celeb-DF and Celeb-DF++ splits share identities and source-video content between training and test, and that custom protocols reusing test videos and identities in training inflate weaker detectors by up to 0.30 AUC.
Intraoperative scene perception in ophthalmic microsurgery is limited by instrument occlusion: when a tool enters the high-magnification field, it conceals tissue from both the surgeon and the perception components of robot-assisted systems that process the video. Generative inpainting can remove the occluder but replaces the missing image content with synthesis, which is unsuitable when visual provenance matters. Many video inpainters also depend on future frames and cannot operate online. We present TransGHOST, a causal reference-guided reconstruction pipeline whose non-generative safe mode operates in real time. Every restored value is obtained by resampling a recorded frame using geometric transforms fitted only to visible regions. The system provides per-pixel confidence and provenance and marks unsupported pixels as abstentions. We evaluate TransGHOST on a new veterinary clinical corpus of canine cataract surgery with pixel-level instrument and anatomy labels. Only the occluded region is scored, and each scoring target is verified to differ from the reference supplied to the system. Safe mode reaches 31.89 dB masked PSNR and 0.915 masked SSIM at 51.7 ms median and 85.5 ms 95th-percentile latency. On a pre-declared held-out cohort using instrument cutouts excluded from segmentation training, it reaches 29.83 dB masked PSNR, while matched-mask evaluation on validation shows a 3.45 dB advantage over LaMa. Region-aware reference selection improves masked PSNR by 3.60 dB over recency when candidate references can themselves be occluded. Reference-age analysis further defines the operating interval for the brief, repeated occlusions targeted by TransGHOST.
Text-guided segmentation models receive an image and a short caption about the lesion, and on public benchmarks the caption improves accuracy considerably. On colonoscopy and breast ultrasound, however, the size, count and location words of each caption were computed from the mask that later scores the prediction, and the chest X-ray captions were written by benchmark annotators. We tested what two published models take from them. On colonoscopy and chest X-ray, it is mostly the lesion's extent. Keeping only the size word or the lung-zone phrase recovers 84 to 98 percent of the improvement over an empty caption, and removing it loses 52 to 77 percent. The rest matters little, and a polyp called a rock is segmented just as well. We call this extent leakage. On breast ultrasound, nearly half the captions also name the diagnosis, which carries part of the benefit. Retraining on three datasets shows that models pick up this habit from their training captions, though only partly on breast ultrasound, and that it weakens their segmentation from the image alone. The leak can be measured without training a model, since caption words alone predict lesion area (coefficient of determination 0.54 to 0.77 on six of seven test sets). Without the test mask, the best method we tested, which estimates extent from the image, reaches 74.59 Dice on colonoscopy against the published 77.23. We propose an evaluation protocol that keeps what a caption says about the image apart from what it copies from the label.
Latent diffusion models decode approximately correct latents, making decoder sensitivity to perturbations relevant to image quality. After whitening, forward diffusion corrupts every direction at the same signal-to-noise ratio (SNR), despite unequal decoder sensitivities. We study an invertible linear layer between the encoder and diffusion process that shifts each direction's log-SNR. Its linear minimum mean-square-error (LMMSE) recovery risk has an exact spectral expression for any whitened latent distribution. Under an energy budget and a floor ensuring invertibility, minimizing decoder-sensitivity-weighted LMMSE risk integrated over the diffusion schedule yields a unique optimal allocation. The layer aligns with the decoder's expected pullback metric and equalizes marginal risk reduction across directions, computed by a deterministic spectral solver. At large budgets, the allocation follows a power law in sensitivity whose exponent is determined by the low-SNR tail of the loss weighting: one half for noise prediction and two thirds for data prediction under the cosine schedule, with a crossover on truncated schedules. We instantiate this principle as RB-LDC using Jacobian-based sensitivity estimates. On ImageNet at 256 x 256 resolution with SD-VAE and VA-VAE, trained RB-LDC improves reconstruction from noisy latents over budget-matched isotropic coding in all four metrics and five coding configurations, averaged over 15 SNRs and three pipeline seeds. It reduces average perturbed FID by 17.1 for VA-VAE block-local coding while preserving clean reconstruction. We also report intermediate SNRs where perturbed FID worsens. With a frozen diffusion model, RB-LDC yields lower FID point estimates in all 12 tested combinations of sampling progress and channel SNR under corruption during sampling.
In open-world recognition, machine learning models not only make predictions on unknown structures ("generalized open-set"), but also continually learn and improve from those predictions over time. Interactive segmentation, despite its access to user feedback, has so far remained confined to the open-set setting. In this work we introduce MEOW, the first framework that turns non-interactive segmentation backbones into interactive open-world learners without updating learnable model parameters. MEOW combines parameter-free transductive reasoning with memory consolidation to accumulate knowledge and thereby improve future performance. Focusing on medical data, we show that MEOW consistently improves its own predictions on classes never seen during training, and transfers zero-shot to unfamiliar domains, including micro-CT data and other imaging modalities.
Farm security in the UK rests on thinner evidence than its cost suggests. We ran a PRISMA guided review: twelve searches of Web of Science and Scopus on 2nd -4th November 2026 returned 1,183 records (761 unique), and grey-literature and hand-searches added 35 assessed sources. We synthesised 96 database records and 24 grey sources narratively. UK peer-reviewed evidence is small (16 records), and the main UK survey of crime and farmer mental health has 80 responses. Cost estimates rest on insurer claims: NFU Mutual put UK rural crime at £41.5 million in 2025, with livestock theft up almost 30% to about £4.5 million and dog attacks on livestock costing £1.95 million. The Home Office states that agricultural crime cannot be separately identified in its central data. We found no independent evaluation of cameras, sensors, trackers or marking; light and sound deterrents lost effect within one to five months in a systematic review of anti-predator interventions; and no study examined how often farm gates are left open or what follows. Camera-trap AI papers report accuracy on ecological datasets, none at a farm gate in the records we found, and in one comparison machine learning detected 41% of wildlife against 81% for analysts. We argue, as an inference from these gaps and not as a demonstrated effect, that continuous gate-level AI-mediated monitoring, as implemented in Fieldmark AI, would be highly beneficial for UK farm security. It would create the farm-level, time-stamped record that no official source holds, verify events that periodic human checks miss, resolve whether foxes, dogs or people were responsible, and give every other security measure something to be tested against. No study has yet shown that it reduces loss, and we set out the evaluation needed to show it.
Automated fruit grading, sorting, and orientation systems require real-time, accurate detection of fruit-level morphological landmarks rather than the fruit boundary alone. This paper presents a custom, fully self-collected and self-annotated dataset of 1506 apple images (1054 training, 301 validation, 151 test) captured under controlled, fixed-camera conditions and annotated for three target classes: fruit body, stem cavity, and calyx. A YOLO11s detector trained with default hyperparameters is adopted as the baseline, and a systematic, one-factor-at-a-time ablation study is conducted across input resolution, data augmentation policy, optimizer configuration, and lightweight architectural enhancement to identify the factors that most improve detection of these fine-grained landmarks under high inter-cultivar color and texture variability. Neither increased input resolution, nor any of six tested augmentation variants, nor manual optimizer tuning improved on the untuned baseline (F1 = 0.9047); only a same-slot architectural substitution of the backbone's C2PSA block with Coordinate Attention (CA) matched the baseline's overall F1 (0.9038) while reducing parameter count by 10.2% and achieving the study's best [email omitted] (0.935) and best stem-cavity F1 (0.8217). A confidence-threshold sweep identified 0.15 as the deployment operating point for this model, and a confirmatory experiment combining the CA-swap architecture with the best-performing augmentation variant showed the two effects do not combine additively, reversing the architectural gain rather than compounding it. The resulting design recommendations --- the CA-swap architecture alone, at a 0.15 confidence threshold --- target real-time apple landmark detection for orientation-aware industrial sorting and alignment systems.
Multi-modal brain tumor segmentation is essential for clinical diagnosis, radiotherapy planning, and longitudinal monitoring. While recent studies have achieved remarkable overall segmentation performance, they rely on overlap-based objectives that inherently favor large regions, causing sub-3 mm micro-lesions to be systematically suppressed during training. Moreover, existing pixel-level segmentation frameworks lack explicit instance-level modeling, leading to unreliable lesion counting that directly impacts clinical staging and therapeutic decisions. To address these challenges, we propose MicroBT, a unified multi-modal framework for micro brain tumor segmentation and counting. Our core insight is that a fundamental geometric property is inherently scale-invariant: regardless of size, each lesion possesses a well-defined centroid. Leveraging this, we first devise a centroid-anchored synergistic paradigm that utilizes predicted centroids as shared anchors to simultaneously condition dense segmentation and instance-level localization. To counteract the tumor size imbalance, we further devise a Volume-Calibrated Loss (VC-Loss) that adaptively amplifies the gradient contribution of micro-lesions, alongside a Size-Aware Curriculum Optimization (SACO) that stabilizes multi-task training by establishing global representations before progressively unlocking fine-grained localization. Extensive experiments demonstrate that MicroBT outperforms state-of-the-art methods, particularly on metastases, where micro-lesion ($\leq 3$ mm) localization precision improves by approximately 7% and counting estimation errors decrease by approximately 26%. The code is available at https://github.com/leonlou921/MicroBT.
Natural fractures (NFs) and drilling-induced tensile fractures (DITFs) in acoustic image logs provide insights into rock mass structure and in situ stress, but manual characterization is time-consuming. Deep learning can automate this task, although expert-labeled data are scarce. We investigate whether adapting a pretrained vision foundation model (VFM) outperforms task-specific models trained from scratch with limited annotations. Using 80 annotated acoustic image samples, we fine-tuned DINOv3, pretrained on satellite imagery, and evaluated it on four boreholes excluded from fine-tuning. The adapted model estimated NF and DITF parameters comparably to human interpretation. Relative to U-Net, YOLOv8, and Swin Transformer models trained from scratch, it yielded fewer false positives and negatives and more accurate fracture counts and orientations. However, its performance advantage generally narrowed as training sample size decreased. We also assessed decoder architecture, fine-tuning strategy, and image enhancement. A progressive upsampling decoder with full fine-tuning performed best among the DINOv3 configurations tested, while fracture-focused image enhancement did not consistently improve accuracy. These results demonstrate the potential of VFM adaptation for fracture characterization in acoustic image logs with limited labeled data and suggest a transferable workflow for other annotation-scarce geoscientific imaging tasks.
Under a fixed frame budget, video understanding with vision-language models (VLMs) depends on whether the sampled frames provide sufficient evidence for the question. Sufficient frame evidence supports direct answering, whereas less informative observations may benefit from reasoning that connects the available cues. We introduce EVAThink, which enhances video understanding through evidence sufficiency-guided training for adaptive thinking. First, Recoverable Evidence Gap Mining constructs evidence-dense and evidence-sparse video views that share the same question and answer under an identical frame budget. It retains pairs where sparse views yield lower direct-answer accuracy than dense views but benefit from explicit reasoning, providing supervision for response-mode selection. Second, Evidence-Guided Two-Stage Reinforcement Learning uses stage-specific rewards to guide training with evidence-based mode preferences and then refine response efficiency, while first-token advantage regularization mitigates mode collapse. Trained on only 6.7K curated samples, EVAThink consistently improves upon the base model across video understanding benchmarks under both 32- and 64-frame budgets, while outperforming strong baselines on average. The code, model, and datasets are available at \href{https://github.com/ChangtiWu/EVAThink}{EVAThink}.
Three-dimensional scene graphs (3DSGs) can provide robots with persistent metric-semantic representations for spatial reasoning, but the existing methods have primarily been developed for indoor and terrestrial environments. In this paper we investigate how a representative outdoor open-vocabulary method transfers to maritime environments and which adaptations are needed. Our experiments show that zero-shot representative outdoor 3DSG descriptions agree with our labeled ground truth for a majority of geometry-supported objects on our data, but the resulting map exhibits considerable physical-entity fragmentation, and its terrestrial higher-level assumptions don't capture maritime spatial organization needs. We therefore keep the underlying mapping method and post-process its accumulated evidence. We show a metric-semantic consolidation method followed by geometry-grounded multi-view vision-language model (VLM) reasoning that reconciles remaining fragments into physical entities and groups entities through observations into local places. In parallel, we construct maritime macro regions using map priors and onboard water and structure observations, and associate them with the local places. The resulting representation reduces physical-entity fragmentation beyond both the baseline and our geometry-semantic consolidation component alone, supports correspondence of fixed local places across repeated maritime traversals, and improves contextual retrieval on our reviewed queries over tested baselines. These results show how an outdoor open-vocabulary map can be adapted into persistent maritime scene memory with our methods for object/place identity and spatial abstractions.
Component-level information supports rooftop photovoltaic (PV) inspection, but recognizing individual components across complex scenes remains challenging. We present PVIS-10K, a dataset of 10,065 drone images with component-level annotations covering 63,632 physical PV modules across 22 rooftop installations. The high-resolution red-green-blue (RGB) images were acquired by unmanned aerial vehicles (UAVs, also known as drones) in Ningbo, Shaoxing and Shenzhen, China. Multi-angle oblique observations capture installation-height and scale variation, dense arrangements, occlusion, shadows and reflections. Three-dimensional delineation, reprojection and image-by-image review produced 3,516,675 image-level polygon annotations, stored in Common Objects in Context (COCO) format. Station-disjoint partitions support evaluation on unseen stations within and across cities. Four supervised models and text-prompted zero-shot Segment Anything Model 3 (SAM 3) provide full-resolution benchmarks of instance matching and foreground coverage. PVIS-10K supports component localization, delineation and image-based counting for drone-based rooftop PV inspection research.
Recent work has shown that pretrained image generators can support diverse visual tasks. Yet directly applying them to language-guided unmanned aerial vehicle (UAV) segmentation exposes a key limitation: a plausible mask may correspond to the wrong instance among small and repeated objects. We call this mismatch between mask rendering and expression-level target selection the Grounding-Rendering Gap. We present DroneSegEdit, which combines RGB proposal rendering with language-grounded, proposal-constrained verification to select supported foreground while retaining rendered geometry. DroneSegEdit adapts the open-weight LongCat-Image-Edit model using task-specific LoRA and fixed color-to-label decoding. For referring segmentation, a Language-Grounded Spatial Verifier (LGSV) reads hierarchical states from the frozen multimodal large language model backbone to derive expression-conditioned spatial evidence, which is mapped back to the rendered proposal without introducing foreground outside it. On RIS-LAD, DroneSegEdit reaches a new state of the art at 52.78 mIoU, 7.16 points above the previous best. It remains competitive overall on UAVid-RIS and VDD-RIS, with its clearest class-wise gains on compact, application-relevant categories, notably vehicles and humans. A separate verifier-free experiment extends the RGB interface to full-scene parsing: a task-specific adapter jointly renders all classes from one task-level instruction without enumerating class names or palette colors.
BraTS-METS 2026 Task 1 evaluates segmentation of brain metastases and postoperative resection cavities on heterogeneous multiparametric MRI, requiring detection of small enhancing foci while preserving the geometry and hierarchy of enhancing tumor (ET), tumor core (TC), whole tumor (WT), and resection cavity (RC). We present COMPASS-METS (Component-Aware Multi-Trainer Probability Assembly with Structured Segmentation), which combines three five-fold nnU-Net configurations into aligned regional probability ensembles. A structured-probability anchor applies region-specific component decisions, TC boundary completion, RC filtering, and hierarchy-preserving conversion. Omitted ET candidates are recovered additively using learned component confidence and deterministic multi-trainer support for ET and its TC/WT parents; a final utility gate admits only disconnected candidates with sufficient lesion-existence and geometry-safety probabilities. On the reused 179-case challenge development cohort, COMPASS-METS (Team: MicroBT) achieved Small F1 0.354, lesion-wise DSC 0.712, NSD 0.714, All F1 0.779, and HD95 56.6 mm. Relative to ResEncXL, it improved DSC, NSD, and All F1 and reduced HD95, while Small F1 changed from 0.367 to 0.354. Stagewise analysis attributes the principal geometric gains to structured probability construction and shows that guarded ET refinement recovers most of the associated small-lesion detection loss. These 179-case results remain selection-aware development estimates. The source code is publicly available at https://github.com/leonlou921/COMPASS-METS.
Computational approaches are increasingly being explored to streamline histopathology workflows, including label-free methods that aim to reduce reliance on chemical staining. Among these, deep learning-based label-free virtual staining uses images of unstained tissue to computationally generate stained images. Despite growing interest, model development and validation remain constrained by the scarcity of paired, clinically annotated whole-slide imaging datasets. We present BRIDGE, a dataset of prostate core-needle biopsies collected at the Prof. Franciszek Lukaszczyk Oncology Centre in Bydgoszcz, Poland. The dataset comprises 869 glass slides from 169 patients, each digitized sequentially after sectioning, deparaffinization, and hematoxylin and eosin staining, yielding in total 2607 whole-slide images. The dataset supports the development and external validation of label-free virtual staining, virtual unstaining, image-registration and segmentation methods. Each slide is accompanied by a Gleason score, enabling downstream analyses of cancer classification and grading concordance. The stained images can be used independently for color normalization and cross-scanner robustness studies. All images are de-identified, distributed in their native scanner format and will be available through the BioImage Archive.
Establishing dense correspondence between high-resolution 3D shapes remains challenging due to the spatial--spectral coupling at the core of many functional-map methods, which grows quadratically with mesh resolution. We observe that this coupling is equivalent to a scaled dot-product attention formulation, with learned descriptors serving as queries and keys and spectral basis functions as values. This connection makes efficient attention implementations directly applicable to shape matching and an effective mechanism for scaling spatial-spectral coupling. We show that optimized quadratic attention implementations can scale modern deep functional map pipelines to large mesh sizes, improving both runtime and memory consumption, while sub-quadratic methods introduce errors, particularly for non-isometric shape pairs, where an exact coupling proves most necessary. These insights motivate \textit{Geo-FlashAttention}, which restricts the exact softmax to geometry-selected surface patches and achieves near-linear empirical scaling while preserving correspondence accuracy of exact implementations. Across six standard benchmarks and three functional-map pipelines, rigorous analysis shows that efficient couplings preserve correspondence accuracy. At $60k$ vertices, these insights reduce peak training memory from 106.08 to 5.67\,GiB, with Geo-FlashAttention being $1.5\times$ faster per training step than FlashAttention. This enables end-to-end learning for correspondence problems on large meshes such as cortical surfaces, where high resolution is required to preserve disease-relevant intricacies of the surface geometry. Project page available https://spatial-spectral-attention.github.io/
Deep neural networks underpin fundamental tasks in a wide range of signal domains, such as image, audio, and language processing. However, their internal mechanisms are notably difficult to interpret. Internal activations are structured signals transformed across layers that may encode a multitude of interdependent relationships, as individual neurons frequently represent multiple underlying concepts. Mechanistic interpretability aims to uncover the internal mechanisms of neural networks by identifying components whose activations are causally responsible for specific behaviors. In such a context, this article provides an overview of advances in mechanistic interpretability for the analysis of causal inference in deep neural networks.