Automated quality assurance is essential for low-dose computed tomography (LDCT) lung screening, yet manual checks strain clinical workflows. We present a fully automated artificial intelligence tool that quantifies scan coverage and image noise in LDCT without user input. Lungs and the aorta are segmented to measure cranial/caudal over- and underscanning, and noise is computed as the standard deviation of Hounsfield units (HUs) within descending aortic blood, normalized to a 1 mm3 voxel. Performance was verified in a reader study of 98 LDCT scans from the National Lung Screening Trial (NLST), and then applied to 38,834 NLST scans reconstructed with a standard kernel. In the reader study, lung masks were rated ≥“Nearly Perfect” in 90.8% and aorta-blood masks in 96.9% of cases. Across 38,834 scans, mean overscanning distances were 31.21 mm caudally and 14.54 mm cranially; underscanning occurred in 4.36% (caudal) and 0.89% (cranial). The tool enables objective, large-scale monitoring of LDCT quality—reducing routine manual workload through exception-based human oversight, flagging protocol deviations, and supporting cross-center benchmarking—and may facilitate dose optimization by reducing systematic over- and underscanning.

Vision foundation models (VFMs) trained on large-scale image datasets provide high-quality features that have significantly advanced 2D visual recognition. However, their potential in 3D scene segmentation remains largely untapped, despite the common availability of 2D images alongside 3D point cloud datasets. While significant research has been dedicated to 2D-3D fusion, recent state-of-the-art 3D methods predominantly focus on 3D data, leaving the integration of VFMs into 3D models underexplored. In this work, we challenge this trend by introducing DITR, a generally applicable approach that extracts 2D foundation model features, projects them to 3D, and finally injects them into a 3D point cloud segmentation model. DITR achieves state-of-the-art results on both indoor and outdoor 3D semantic segmentation benchmarks. To enable the use of VFMs even when images are unavailable during inference, we additionally propose to pretrain 3D models by distilling 2D foundation models. By initializing the 3D backbone with knowledge distilled from 2D VFMs, we create a strong basis for downstream 3D segmentation tasks, ultimately boosting performance across various datasets.

Despite the advances in automated medical image segmentation, AI models still underperform in various clinical settings, posing challenges for integration into real-world workflows. In this pre-registered prospective multicenter evaluation, we analyzed 20 state-of-the-art mandibular segmentation models across 19,218 segmentations of 1,000 clinically resampled CT/CBCT scans. Our results suggest that for a given model, segmentation accuracy can vary by up to 25% in Dice score as socio-technical factors such as voxel size, bone orientation, and patient conditions (e.g., osteosynthesis or pathology) shift from favorable to adverse. Higher sharpness, isotropic smaller voxels, and neutral orientation significantly improved results, while metallic osteosynthesis and anatomical complexity led to significant degradation. Our findings challenge the common view of AI models as “plug-and-play” tools and suggest evidence-based optimization recommendations for both clinicians and developers. This will in turn boost the integration of AI segmentation tools in routine healthcare.

Developing a deep-learning model for automated multi-tissue, multi-condition knee MRI analysis and assessing its clinical potential. This retrospective dual-center study included 3121 MRI studies from 3018 adults, who underwent routine knee MRI examinations at a radiologic practice (2012–2019). Twenty-three conditions across cartilage, menisci, bone marrow, ligaments, and other soft tissues were manually labeled. A 3D slice transformer network was trained for binary classification and evaluated in terms of the area under the receiver operating characteristic curve (AUC), sensitivity, and specificity using a five-fold cross-validation and an external test set of 448 MRI studies (429 adults) from a university hospital (2022–2023). To assess differences in diagnostic performance, two inexperienced and two experienced radiology residents read 50 external test studies with and without model assistance. Paired t-tests were used for statistical analysis. Averaged over cross-validation tests, the model’s AUC was at least 0.85 for 8 conditions and at least 0.75 for 18 conditions. Generalization on the external test set was robust, with a mean absolute AUC difference of 0.05 ± 0.03 per condition. Model assistance improved accuracy and sensitivity for inexperienced residents, increased inter-reader agreement for both groups, and increased sensitivity and shortened reading times by 10

Recent work showed that large diffusion models can be reused as highly precise monocular depth estimators by casting depth estimation as an image-conditional image generation task. While the proposed model achieved state-of-the-art results, high computational demands due to multi-step inference limited its use in many scenarios. In this paper, we show that the perceived inefficiency was caused by a flaw in the inference pipeline that has so far gone unnoticed. The fixed model performs comparably to the best previously reported configuration while being more than 200x faster. To optimize for downstream task performance, we perform end-to-end fine-tuning on top of the single-step model with task-specific losses and get a deterministic model that outperforms all other diffusion-based depth and normal estimation models on common zero-shot benchmarks. We surprisingly find that this fine-tuning protocol also works directly on Stable Diffusion and achieves comparable performance to current state-of-the-art diffusion-based depth and normal estimation models, calling into question some of the conclusions drawn from prior works.

Safe navigation of self-driving cars and robots requires a precise understanding of their environment. Training data for perception systems cannot cover the wide variety of objects that may appear during deployment. Thus, reliable identification of unknown objects, such as wild animals and untypical obstacles, is critical due to their potential to cause serious accidents. Significant progress in semantic segmentation of anomalies has been facilitated by the availability of out-of-distribution (OOD) benchmarks. However, a comprehensive understanding of scene dynamics requires the segmentation of individual objects, and thus the segmentation of instances is essential. Development in this area has been lagging, largely due to the lack of dedicated benchmarks. The situation is similar in object detection. While there is interest in detecting and potentially tracking every anomalous object, the availability of dedicated benchmarks is clearly limited. To address this gap, this work extends some commonly used anomaly segmentation benchmarks to include the instance segmentation and object detection tasks. Our evaluation of anomaly instance segmentation and object detection methods shows that both of these challenges remain unsolved problems. We provide a competition and benchmark website under https://vision.rwth-aachen.de/oodis

Research into Video Large Language Models (LLMs) has progressed rapidly, with numerous models and benchmarks emerging in just a few years. Typically, these models are initialized with a pretrained text-only LLM and finetuned on both image- and video-caption datasets. In this paper, we present findings indicating that Video LLMs are more capable of temporal reasoning after image-only training than one would assume, and that improvements from video-specific training are surprisingly small. Specifically, we show that image-trained versions of two LLMs trained with the recent LongVU algorithm perform significantly above chance level on TVBench, a temporal reasoning benchmark. Additionally, we introduce a simple finetuning scheme involving sequences of annotated images and questions targeting temporal capabilities. This baseline results in temporal reasoning performance close to, and occasionally higher than, what is achieved by video-trained LLMs. This suggests suboptimal utilization of rich temporal features found in real video by current models. Our analysis motivates further research into the mechanisms that allow image-trained LLMs to perform temporal reasoning, as well as into the bottlenecks that render current video training schemes inefficient.

Sa2VA is a recent model for language-guided dense grounding in images and video that achieves state-of-the-art results on multiple segmentation benchmarks and that has become widely popular. However, we found that Sa2VA does not perform according to its full potential for referring video object segmentation tasks. We identify inconsistencies between training and inference procedures as the key factor holding it back. To mitigate this issue, we propose an improved version of Sa2VA, Sa2VA-i, that rectifies these issues and improves the results. In fact, Sa2VA-i sets a new state of the art for multiple video benchmarks and achieves improvements of up to +11.6 J F on MeViS, +1.4 on Ref-YT-VOS, +3.3 on Ref-DAVIS and +4.1 on ReVOS using the same Sa2VA checkpoints. With our fixes, the Sa2VA-i-1B model even performs on par with the original Sa2VA-26B model on the MeViS benchmark. We hope that this work will show the importance of seemingly trivial implementation details and that it will provide valuable insights for the referring video segmentation field. We provide the code and updated models at https://github.com/kumuji/sa2va-i

3D Gaussian Splatting (3DGS) has emerged as a powerful representation for neural scene reconstruction, offering high-quality novel view synthesis while maintaining computational efficiency. In this paper, we extend the capabilities of 3DGS beyond pure scene representation by introducing an approach for open-vocabulary 3D instance segmentation without requiring manual labeling, termed OpenSplat3D. Our method leverages feature-splatting techniques to associate semantic information with individual Gaussians, enabling fine-grained scene understanding. We incorporate Segment Anything Model instance masks with a contrastive loss formulation as guidance for the instance features to achieve accurate instance-level segmentation. Furthermore, we utilize language embeddings of a vision-language model, allowing for flexible, text-driven instance identification. This combination enables our system to identify and segment arbitrary objects in 3D scenes based on natural language descriptions. We show results on LERF-mask and LERF-OVS as well as the full ScanNet++ validation set, demonstrating the effectiveness of our approach.

The detection and classification of exfoliated two-dimensional (2D) material flakes from optical microscope images can be automated using computer vision algorithms. This has the potential to increase the accuracy and objectivity of classification and the efficiency of sample fabrication, and it allows for large-scale data collection. Existing algorithms often exhibit challenges in identifying low-contrast materials and typically require large amounts of training data. Here, we present a deep learning model, called MaskTerial, that uses an instance segmentation network to reliably identify 2D material flakes. The model is extensively pre-trained using a synthetic data generator that generates realistic microscopy images from unlabeled data. This results in a model that can quickly adapt to new materials with as little as 5 to 10 images. Furthermore, an uncertainty estimation model is used to finally classify the predictions based on optical contrast. We evaluate our method on eight different datasets comprising five different 2D materials and demonstrate significant improvements over existing techniques in the detection of low-contrast materials such as hexagonal boron nitride.

Vision Transformers (ViTs) have shown remarkable performance and scalability across various computer vision tasks. To apply single-scale ViTs to image segmentation, existing methods adopt a convolutional adapter to generate multi-scale features, a pixel decoder to fuse these features, and a Transformer decoder that uses the fused features to make predictions. In this paper, we show that the inductive biases introduced by these task-specific components can instead be learned by the ViT itself, given sufficiently large models and extensive pre-training. Based on these findings, we introduce the Encoder-only Mask Transformer (EoMT), which repurposes the plain ViT architecture to conduct image segmentation. With large-scale models and pre-training, EoMT obtains a segmentation accuracy similar to state-of-the-art models that use task-specific components. At the same time, EoMT is significantly faster than these methods due to its architectural simplicity, e.g., up to 4x faster with ViT-L. Across a range of model sizes, EoMT demonstrates an optimal balance between segmentation accuracy and prediction speed, suggesting that compute resources are better spent on scaling the ViT itself rather than adding architectural complexity. Code: https://www.tue-mps.org/eomt/.

This report presents an overview of the 7th Large-scale Video Object Segmentation (LSVOS) Challenge held in conjunction with ICCV 2025. Besides the two traditional tracks of LSVOS that jointly target robustness in realistic video scenarios: Classic VOS (VOS), and Referring VOS (RVOS), the 2025 edition features a newly introduced track, Complex VOS (MOSEv2). Building upon prior insights, MOSEv2 substantially increases difficulty, introducing more challenging but realistic scenarios including denser small objects, frequent disappear/reappear events, severe occlusions, adverse weather and lighting, etc., pushing long-term consistency and generalization beyond curated benchmarks. The challenge retains standard J, F, and J&F metrics for VOS and RVOS, while MOSEv2 adopts J&Ḟ as the primary ranking metric to better evaluate objects across scales and disappearance cases. We summarize datasets and protocols, highlight top-performing solutions, and distill emerging trends, such as the growing role of LLM/MLLM components and memory-aware propagation, aiming to chart future directions for resilient, language-aware video segmentation in the wild.

In this paper we create an RGB-D 3D object detector targeted at indoor robotics use cases where one modality may be unavailable due to a specific sensor setup or a sensor failure. We incorporate RGB and depth fusion into the recent Cube R-CNN framework with support for selective modality dropout. To train this model, we augment the Omni3DIN dataset with depth information, leading to a diverse dataset for 3D object detection in indoor scenes. In order to leverage strong pretrained networks, we investigate the viability of Transformer-based backbones (Swin, ViT) as an alternative to the currently popular CNN-based DLA backbone. We show that these Transformer-based image models work well based on our early-fusion approach and propose a modality dropout scheme to avoid the disregard of any modality during training, facilitating selective modality dropout during inference. In extensive experiments, our proposed RGB-D Cube R-CNN outperforms an RGB-only Cube R-CNN baseline by a significant margin on the task of indoor object detection. Additionally, we observe a slight performance boost from the RGB-D training when inferring on only one modality, which could for example be valuable in robotics applications with a reduced or unreliable sensor set. Code and scripts to recreate the RGB-D dataset can be found at: https://github.com/VisualComputingInstitute/omni3d-rgbd

Objectives Tooth extraction is one of the most frequently performed medical procedures. The indication is based on the combination of clinical and radiological examination and individual patient parameters and should be made with great care. However, determining whether a tooth should be extracted is not always a straightforward decision. Moreover, visual and cognitive pitfalls in the analysis of radiographs may lead to incorrect decisions. Artificial intelligence (AI) could be used as a decision support tool to provide a score of tooth extractability.Material and methods Using 26,956 single teeth images from 1,184 panoramic radiographs (PANs), we trained a ResNet50 network to classify teeth as either extraction-worthy or preservable. For this purpose, teeth were cropped with different margins from PANs and annotated. The usefulness of the AI-based classification as well that of dentists was evaluated on a test dataset. In addition, the explainability of the best AI model was visualized via a class activation mapping using CAMERAS.Results The ROC-AUC for the best AI model to discriminate teeth worthy of preservation was 0.901 with 2% margin on dental images. In contrast, the average ROC-AUC for dentists was only 0.797. With a 19.1% tooth extractions prevalence, the AI model's PR-AUC was 0.749, while the dentist evaluation only reached 0.589.Conclusion AI models outperform dentists/specialists in predicting tooth extraction based solely on X-ray images, while the AI performance improves with increasing contextual information.Clinical relevance AI could help monitor at-risk teeth and reduce errors in indications for extractions.

Recently, the self-supervised learning framework data2vec has shown inspiring performance for various modalities using a masked student-teacher approach. However, it remains open whether such a framework generalizes to the unique challenges of 3D point clouds. To answer this question, we extend data2vec to the point cloud domain and report encouraging results on several downstream tasks. In an in-depth analysis, we discover that the leakage of positional information reveals the overall object shape to the student even under heavy masking and thus hampers data2vec to learn strong representations for point clouds. We address this 3D-specific shortcoming by proposing point2vec, which unleashes the full potential of data2vec-like pre-training on point clouds. Our experiments show that point2vec outperforms other self-supervised methods on shape classification and few-shot learning on ModelNet40 and ScanObjectNN, while achieving competitive results on part segmentation on ShapeNetParts. These results suggest that the learned representations are strong and transferable, highlighting point2vec as a promising direction for self-supervised learning of point cloud representations.

This study investigates the application of ordinal regression methods for categorizing disease severity in chest radiographs. We propose a framework that divides the ordinal regression problem into three parts: a model, a target function, and a classification function. Different encoding methods, including one-hot, Gaussian, progress-bar, and our soft-progress-bar, are applied using ResNet50 and ViT-B-16 deep learning models. We show that the choice of encoding has a strong impact on performance and that the best encoding depends on the chosen weighting of Cohen's kappa and also on the model architecture used. We make our code publicly available on GitHub.

A single unexpected object on the road can cause an accident or may lead to injuries. To prevent this, we need a reliable mechanism for finding anomalous objects on the road. This task, called anomaly segmentation, can be a stepping stone to safe and reliable autonomous driving. Current approaches tackle anomaly segmentation by assigning an anomaly score to each pixel and by grouping anomalous regions using simple heuristics. However, pixel grouping is a limiting factor when it comes to evaluating the segmentation performance of individual anomalous objects. To address the issue of grouping multiple anomaly instances into one, we propose an approach that produces accurate anomaly instance masks. Our approach centers on an out-of-distribution segmentation model for identifying uncertain regions and a strong generalist segmentation model for anomaly instances segmentation. We investigate ways to use uncertain regions to guide such a segmentation model to perform segmentation of anomalous instances. By incorporating strong object priors from a generalist model we additionally improve the per-pixel anomaly segmentation performance. Our approach outperforms current pixel-level anomaly segmentation methods, achieving an AP of 80.08% and 88.98% on the Fishyscapes Lost and Found and the RoadAnomaly validation sets respectively. Project page: https://vision.rwth-aachen.de/ugains

Cell segmentation in bright-field histological slides is a crucial topic in medical image analysis. Having access to accurate segmentation allows researchers to examine the relationship between cellular morphology and clinical observations. Unfortunately, most segmentation methods known today are limited to nuclei and cannot segment the cytoplasm. We present a new network architecture Cyto R-CNN that is able to accurately segment whole cells (with both the nucleus and the cytoplasm) in bright-field images. We also present a new dataset CytoNuke, consisting of multiple thousand manual annotations of head and neck squamous cell carcinoma cells. Utilizing this dataset, we compared the performance of Cyto R-CNN to other popular cell segmentation algorithms, including QuPath’s built-in algorithm, StarDist, Cellpose and a multi-scale Attention Deeplabv3+. To evaluate segmentation performance, we calculated AP50, AP75 and measured 17 morphological and staining-related features for all detected cells. We compared these measurements to the gold standard of manual segmentation using the Kolmogorov–Smirnov test. Cyto R-CNN achieved an AP50 of 58 . 65 % and an AP75 of 11 . 56 % in whole-cell segmentation, outperforming all other methods (QuPath 19 . 46 / 0 . 91 %; StarDist 45 . 33 / 2 . 32 %; Cellpose 31 . 85 / 5 . 61 %, Deeplabv3+ 3 . 97 / 1 . 01 %). Cell features derived from Cyto R-CNN showed the best agreement to the gold standard (D ¯ = 0 . 15) outperforming QuPath (D ¯ = 0 . 22), StarDist (D ¯ = 0 . 25), Cellpose (D ¯ = 0 . 23) and Deeplabv3+ (D ¯ = 0 . 33). Our newly proposed Cyto R-CNN architecture outperforms current algorithms in whole-cell segmentation while providing more reliable cell measurements than any other model. This could improve digital pathology workflows, potentially leading to improved diagnosis. Moreover, our published dataset can be used to develop further models in the future.