arXiv subjects

Klara Janouskova

Publications and source records attributed to Klara Janouskova.

11 recordsLinked to original sources

arXiv2026

Doomed to Re-Annotate, Forever: The ImageNet Story

Top-1 accuracy on ImageNet-1k remains the most commonly reported metric in visual recognition. Quality issues with the dataset have been repeatedly reported, yet the original 2012 noisy labels are still predominantly used. The paper presents a comprehensive effort, which goes well beyond prior correction attempts, towards obtaining accurate and complete ImageNet-1k validation set annotations. The result, ReImageNet, includes multilabel correction, object localization, revised class definitions, and semantic attributes (text-recognition, rendition, reflection, crowd, dominant). The reannotation reveals that approximately 12% of the original ImageNet-1k labels are incorrect, 33.3% of images are multilabel and 3.8% contain no object from an ImageNet-1k class. With the new labels, top-1 accuracy increases by up to 1.2% for supervised models and by 5-6% for MLLMs. We argue that annotation at ImageNet scale cannot realistically be completed in one pass, as errors and definitional issues are discovered only through annotating, and we build our pipeline around repeated refinement and error checking. We observed that human and LLM collaboration with appropriate tooling represents the current quality ceiling for annotation at this scale. ImageNet-1k issues propagate into its derivative test sets, indicating that the problem is structural rather than specific to any single benchmark. All annotations, class definitions, guidelines, and analysis code have been publicly released. Project page: https://vrg.fel.cvut.cz/reimagenet Annotations: https://huggingface.co/datasets/vrg-prague/ReImageNet Code: https://github.com/klarajanouskova/ImageNet

cs.CV↗

arXiv2026

Multimodal Large Language Models as Image Classifiers

Multimodal Large Language Models (MLLM) classification performance depends critically on evaluation protocol and ground truth quality. Studies comparing MLLMs with supervised and vision-language models report conflicting conclusions, and we show these conflicts stem from protocols that either inflate or underestimate performance. Across the most common evaluation protocols, we identify and fix key issues: model outputs that fall outside the provided class list and are discarded, inflated results from weak multiple-choice distractors, and an open-world setting that underperforms only due to poor output mapping. We additionally quantify the impact of commonly overlooked design choices - batch size, image ordering, and text encoder selection - showing they substantially affect accuracy. Evaluating on ReGT, our multilabel reannotation of 625 ImageNet-1k classes, reveals that MLLMs benefit most from corrected labels (up to +10.8%), substantially narrowing the perceived gap with supervised models. Much of the reported MLLMs underperformance on classification is thus an artifact of noisy ground truth and flawed evaluation protocol rather than genuine model deficiency. Models less reliant on supervised training signals prove most sensitive to annotation quality. Finally, we show that MLLMs can assist human annotators: in a controlled case study, annotators confirmed or integrated MLLMs predictions in approximately 50% of difficult cases, demonstrating their potential for large-scale dataset curation. This work is part of the Aiming for Perfect ImageNet-1k project, see https://klarajanouskova.github.io/ImageNet/.

cs.CV↗

arXiv2026

Koo-Fu CLIP: Closed-Form Adaptation of Vision-Language Models via Fukunaga-Koontz Linear Discriminant Analysis

Visual-language models such as CLIP provide powerful general-purpose representations, but their raw embeddings are not optimized for supervised classification, often exhibiting limited class separation and excessive dimensionality. We propose Koo-Fu CLIP, a supervised CLIP adaptation method based on Fukunaga-Koontz Linear Discriminant Analysis, which operates in a whitened embedding space to suppress within-class variation and enhance between-class discrimination. The resulting closed-form linear projection reshapes the geometry of CLIP embeddings, improving class separability while performing effective dimensionality reduction, and provides a lightweight and efficient adaptation of CLIP representations. Across large-scale ImageNet benchmarks, nearest visual prototype classification in the Koo-Fu CLIP space improves top-1 accuracy from 75.1% to 79.1% on ImageNet-1K, with consistent gains persisting as the label space expands to 14K and 21K classes. The method supports substantial compression by up to 10-12x with little or no loss in accuracy, enabling efficient large-scale classification and retrieval.

cs.CV↗

arXiv2025

Robust Context-Aware Object Recognition

In visual recognition, both the object of interest (referred to as foreground, FG, for simplicity) and its surrounding context (background, BG) play an important role. However, standard supervised learning often leads to unintended over-reliance on the BG, known as shortcut learning of spurious correlations, limiting model robustness in real-world deployment settings. In the literature, the problem is mainly addressed by suppressing the BG, sacrificing context information for improved generalization. We propose RCOR -- Robust Context-Aware Object Recognition -- the first approach that jointly achieves robustness and context-awareness without compromising either. RCOR treats localization as an integral part of recognition to decouple object-centric and context-aware modelling, followed by a robust, non-parametric fusion. It improves the performance of both supervised models and VLM on datasets with both in-domain and out-of-domain BG, even without fine-tuning. The results confirm that localization before recognition is now possible even in complex scenes as in ImageNet-1k.

cs.CV↗

arXiv2025

Image Recognition with Vision and Language Embeddings of VLMs

Vision-language models (VLMs) have enabled strong zero-shot classification through image-text alignment. Yet, their purely visual inference capabilities remain under-explored. In this work, we conduct a comprehensive evaluation of both language-guided and vision-only image classification with a diverse set of dual-encoder VLMs, including both well-established and recent models such as SigLIP 2 and RADIOv2.5. The performance is compared in a standard setup on the ImageNet-1k validation set and its label-corrected variant. The key factors affecting accuracy are analysed, including prompt design, class diversity, the number of neighbours in k-NN, and reference set size. We show that language and vision offer complementary strengths, with some classes favouring textual prompts and others better handled by visual similarity. To exploit this complementarity, we introduce a simple, learning-free fusion method based on per-class precision that improves classification performance. The code is available at: https://github.com/gonikisgo/bmvc2025-vlm-image-recognition.

cs.CV↗

arXiv2025

SAM2RL: Towards Reinforcement Learning Memory Control in Segment Anything Model 2

Segment Anything Model 2 (SAM 2) has demonstrated strong performance in object segmentation tasks and has become the state-of-the-art for visual object tracking. The model stores information from previous frames in a memory bank, enabling temporal consistency across video sequences. Recent methods augment SAM 2 with hand-crafted update rules to better handle distractors, occlusions, and object motion. We propose a fundamentally different approach using reinforcement learning for optimizing memory updates in SAM 2 by framing memory control as a sequential decision-making problem. In an overfitting setup with a separate agent per video, our method achieves a relative improvement over SAM 2 that exceeds by more than three times the gains of existing heuristics. These results reveal the untapped potential of the memory bank and highlight reinforcement learning as a powerful alternative to hand-crafted update rules for memory control in visual object tracking.

cs.CV↗

arXiv2025

FungiTastic: A multi-modal dataset and benchmark for image categorization

We introduce a new, challenging benchmark and a dataset, FungiTastic, based on fungal records continuously collected over a twenty-year span. The dataset is labelled and curated by experts and consists of about 350k multimodal observations of 6k fine-grained categories (species). The fungi observations include photographs and additional data, e.g., meteorological and climatic data, satellite images, and body part segmentation masks. FungiTastic is one of the few benchmarks that include a test set with DNA-sequenced ground truth of unprecedented label reliability. The benchmark is designed to support (i) standard closed-set classification, (ii) open-set classification, (iii) multi-modal classification, (iv) few-shot learning, (v) domain shift, and many more. We provide tailored baselines for many use cases, a multitude of ready-to-use pre-trained models on https://huggingface.co/collections/BVRA/fungitastic-66a227ce0520be533dc6403b, and a framework for model training. The documentation and the baselines are available at https://github.com/BohemianVRA/FungiTastic/ and https://www.kaggle.com/datasets/picekl/fungitastic.

cs.CV↗

arXiv2025

Bringing the Context Back into Object Recognition, Robustly

In object recognition, both the subject of interest (referred to as foreground, FG, for simplicity) and its surrounding context (background, BG) may play an important role. However, standard supervised learning often leads to unintended over-reliance on the BG, limiting model robustness in real-world deployment settings. The problem is mainly addressed by suppressing the BG, sacrificing context information for improved generalization. We propose "Localize to Recognize Robustly" (L2R2), a novel recognition approach which exploits the benefits of context-aware classification while maintaining robustness to distribution shifts. L2R2 leverages advances in zero-shot detection to localize the FG before recognition. It improves the performance of both standard recognition with supervised training, as well as multimodal zero-shot recognition with VLMs, while being robust to long-tail BGs and distribution shifts. The results confirm localization before recognition is possible for a wide range of datasets and they highlight the limits of object detection on others

cs.CV↗

arXiv2024

Flaws of ImageNet, Computer Vision's Favourite Dataset

Since its release, ImageNet-1k dataset has become a gold standard for evaluating model performance. It has served as the foundation for numerous other datasets and training tasks in computer vision. As models have improved in accuracy, issues related to label correctness have become increasingly apparent. In this blog post, we analyze the issues in the ImageNet-1k dataset, including incorrect labels, overlapping or ambiguous class definitions, training-evaluation domain shifts, and image duplicates. The solutions for some problems are straightforward. For others, we hope to start a broader conversation about refining this influential dataset to better serve future research.

cs.CV↗

arXiv2024

Single Image Test-Time Adaptation for Segmentation

Test-Time Adaptation (TTA) methods improve the robustness of deep neural networks to domain shift on a variety of tasks such as image classification or segmentation. This work explores adapting segmentation models to a single unlabelled image with no other data available at test-time. In particular, this work focuses on adaptation by optimizing self-supervised losses at test-time. Multiple baselines based on different principles are evaluated under diverse conditions and a novel adversarial training is introduced for adaptation with mask refinement. Our additions to the baselines result in a 3.51 and 3.28 % increase over non-adapted baselines, without these improvements, the increase would be 1.7 and 2.16 % only.

cs.CV↗

arXiv2022

Model-Assisted Labeling via Explainability for Visual Inspection of Civil Infrastructures

Labeling images for visual segmentation is a time-consuming task which can be costly, particularly in application domains where labels have to be provided by specialized expert annotators, such as civil engineering. In this paper, we propose to use attribution methods to harness the valuable interactions between expert annotators and the data to be annotated in the case of defect segmentation for visual inspection of civil infrastructures. Concretely, a classifier is trained to detect defects and coupled with an attribution-based method and adversarial climbing to generate and refine segmentation masks corresponding to the classification outputs. These are used within an assisted labeling framework where the annotators can interact with them as proposal segmentation masks by deciding to accept, reject or modify them, and interactions are logged as weak labels to further refine the classifier. Applied on a real-world dataset resulting from the automated visual inspection of bridges, our proposed method is able to save more than 50\% of annotators' time when compared to manual annotation of defects.

cs.CV↗