Recent studies have underscored the capabilities of natural imaging foundation models to serve as powerful feature extractors, even in a zero-shot setting for medical imaging data. Most commonly, a shallow multi-layer perceptron (MLP) is appended to the feature extractor to facilitate end-to-end learning and downstream prediction tasks such as classification, thus representing the de facto standard. However, as graph neural networks (GNNs) have become a practicable choice for various tasks in medical research in the recent past, we direct attention to the question of how effective GNNs are compared to MLP prediction heads for the task of 3D medical image classification, proposing them as a potential alternative. In our experiments, we devise a subject-level graph for each volumetric dataset instance. Therein latent representations of all slices in the volume, encoded through a DINOv2 pretrained vision transformer (ViT), constitute the nodes and their respective node features. We use public datasets to compare the classification heads numerically and evaluate various graph construction and graph convolution methods in our experiments. Our findings show enhancements of the GNN in classification performance and substantial improvements in runtime compared to an MLP prediction head. Additional robustness evaluations further validate the promising performance of the GNN, promoting them as a suitable alternative to traditional MLP classification heads. Our code is publicly available at: https://github.com/compai-lab/2024-miccai-grail-kiechle
Inspired by human visual attention, deep neural networks have widely adopted attention mechanisms to learn locally discriminative attributes for challenging visual classification tasks. However, existing approaches primarily emphasize the representation of such features while neglecting their precise localization, which often leads to misclassification caused by shortcut biases. This limitation becomes even more pronounced when models are evaluated on transfer or out-of-distribution datasets. In contrast, humans are capable of leveraging prior object knowledge to quickly localize and compare fine-grained attributes, a capability that is especially crucial in complex and high-variance classification scenarios. Motivated by this, we introduce Gaze-CIFAR-10, a human gaze time-series dataset, along with a dual-sequence gaze encoder that models the precise sequential localization of human attention on distinct local attributes. In parallel, a Vision Transformer (ViT) is employed to learn the sequential representation of image content. Through cross-modal fusion, our framework integrates human gaze priors with machine-derived visual sequences, effectively correcting inaccurate localization in image feature representations. Extensive qualitative and quantitative experiments demonstrate that gaze-guided cognitive cues significantly enhance classification accuracy.
Introduction: Photogrammetric surface scans provide a radiation-free option to assess and classify craniosynostosis. Due to the low prevalence of craniosynostosis and high patient restrictions, clinical data is rare. Synthetic data could support or even replace clinical data for the classification of craniosynostosis, but this has never been studied systematically. Methods: We test the combinations of three different synthetic data sources: a statistical shape model (SSM), a generative adversarial network (GAN), and image-based principal component analysis for a convolutional neural network (CNN)-based classification of craniosynostosis. The CNN is trained only on synthetic data, but validated and tested on clinical data. Results: The combination of a SSM and a GAN achieved an accuracy of more than 0.96 and a F1-score of more than 0.95 on the unseen test set. The difference to training on clinical data was smaller than 0.01. Including a second image modality improved classification performance for all data sources. Conclusion: Without a single clinical training sample, a CNN was able to classify head deformities as accurate as if it was trained on clinical data. Using multiple data sources was key for a good classification based on synthetic data alone. Synthetic data might play an important future role in the assessment of craniosynostosis.
The adversarial patch attack against image classification models aims to inject adversarially crafted pixels within a restricted image region (i.e., a patch) for inducing model misclassification. This attack can be realized in the physical world by printing and attaching the patch to the victim object; thus, it imposes a real-world threat to computer vision systems. To counter this threat, we design PatchCleanser as a certifiably robust defense against adversarial patches. In PatchCleanser, we perform two rounds of pixel masking on the input image to neutralize the effect of the adversarial patch. This image-space operation makes PatchCleanser compatible with any state-of-the-art image classifier for achieving high accuracy. Furthermore, we can prove that PatchCleanser will always predict the correct class labels on certain images against any adaptive white-box attacker within our threat model, achieving certified robustness. We extensively evaluate PatchCleanser on the ImageNet, ImageNette, CIFAR-10, CIFAR-100, SVHN, and Flowers-102 datasets and demonstrate that our defense achieves similar clean accuracy as state-of-the-art classification models and also significantly improves certified robustness from prior works. Remarkably, PatchCleanser achieves 83.9% top-1 clean accuracy and 62.1% top-1 certified robust accuracy against a 2%-pixel square patch anywhere on the image for the 1000-class ImageNet dataset.
Existing explanation methods for image classification struggle to provide faithful and plausible explanations. This paper addresses this issue by proposing a post-hoc natural language explanation method that can be applied to any CNN-based classifier without altering its training process or affecting predictive performance. By analysing influential neurons and the corresponding activation maps, the method generates a faithful description of the classifier's decision process in the form of a structured meaning representation, which is then converted into text by a language model. Through this pipeline approach, the generated explanations are grounded in the neural network architecture, providing accurate insight into the classification process while remaining accessible to non-experts. Experimental results show that the NLEs constructed by our method are significantly more plausible and faithful. In particular, user interventions in the neural network structure (masking of neurons) are three times more effective than the baselines.
Early detection of cervical cancer is crucial for improving patient outcomes and reducing mortality by identifying precancerous lesions as soon as possible. As a result, the use of pap smear screening has significantly increased, leading to a growing demand for automated tools that can assist cytologists managing their rising workload. To address this, the Pap Smear Cell Classification Challenge (PS3C) has been organized in association with ISBI in 2025. This project aims to promote the development of automated tools for pap smear images classification. The analyzed images are grouped into four categories: healthy, unhealthy, both, and rubbish images which are considered as unsuitable for diagnosis. In this work, we propose a two-stage ensemble approach: first, a neural network determines whether an image is rubbish or not. If not, a second neural network classifies the image as containing a healthy cell, an unhealthy cell, or both.
Learning accurate classifiers for novel categories from very few examples, known as few-shot image classification, is a challenging task in statistical machine learning and computer vision. The performance in few-shot classification suffers from the bias in the estimation of classifier parameters; however, an effective underlying bias reduction technique that could alleviate this issue in training few-shot classifiers has been overlooked. In this work, we demonstrate the effectiveness of Firth bias reduction in few-shot classification. Theoretically, Firth bias reduction removes the $O(N^{-1})$ first order term from the small-sample bias of the Maximum Likelihood Estimator. Here we show that the general Firth bias reduction technique simplifies to encouraging uniform class assignment probabilities for multinomial logistic classification, and almost has the same effect in cosine classifiers. We derive an easy-to-implement optimization objective for Firth penalized multinomial logistic and cosine classifiers, which is equivalent to penalizing the cross-entropy loss with a KL-divergence between the uniform label distribution and the predictions. Then, we empirically evaluate that it is consistently effective across the board for few-shot image classification, regardless of (1) the feature representations from different backbones, (2) the number of samples per class, and (3) the number of classes. Finally, we show the robustness of Firth bias reduction, in the case of imbalanced data distribution. Our implementation is available at https://github.com/ehsansaleh/firth_bias_reduction
Model compression can significantly reduce the sizes of deep neural network (DNN) models, and thus facilitates the dissemination of sophisticated, sizable DNN models, especially for their deployment on mobile or embedded devices. However, the prediction results of compressed models may deviate from those of their original models. To help developers thoroughly understand the impact of model compression, it is essential to test these models to find those deviated behaviors before dissemination. However, this is a non-trivial task because the architectures and gradients of compressed models are usually not available. To this end, we propose DFLARE, a novel, search-based, black-box testing technique to automatically find triggering inputs that result in deviated behaviors in image classification tasks. DFLARE iteratively applies a series of mutation operations to a given seed image, until a triggering input is found. For better efficacy and efficiency, DFLARE models the search problem as Markov Chains and leverages the Metropolis-Hasting algorithm to guide the selection of mutation operators in each iteration. Further, DFLARE utilizes a novel fitness function to prioritize the mutated inputs that either cause large differences between two models' outputs, or trigger previously unobserved models' probability vectors. We evaluated DFLARE on 21 compressed models for image classification tasks with three datasets. The results show that DFLARE outperforms the baseline in terms of efficacy and efficiency. We also demonstrated that the triggering inputs found by DFLARE can be used to repair up to 48.48% deviated behaviors in image classification tasks and further decrease the effectiveness of DFLARE on the repaired models.
As data volumes continue to grow, the labelling process increasingly becomes a bottleneck, creating demand for methods that leverage information from unlabelled data. Impressive results have been achieved in semi-supervised learning (SSL) for image classification, nearing fully supervised performance, with only a fraction of the data labelled. In this work, we propose a probabilistically principled general approach to SSL that considers the distribution over label predictions, for labels of different complexity, from "one-hot" vectors to binary vectors and images. Our method regularises an underlying supervised model, using a normalising flow that learns the posterior distribution over predictions for labelled data, to serve as a prior over the predictions on unlabelled data. We demonstrate the general applicability of this approach on a range of computer vision tasks with varying output complexity: classification, attribute prediction and image-to-image translation.
In this work, we propose a few-shot colorectal tissue image generation method for addressing the scarcity of histopathological training data for rare cancer tissues. Our few-shot generation method, named XM-GAN, takes one base and a pair of reference tissue images as input and generates high-quality yet diverse images. Within our XM-GAN, a novel controllable fusion block densely aggregates local regions of reference images based on their similarity to those in the base image, resulting in locally consistent features. To the best of our knowledge, we are the first to investigate few-shot generation in colorectal tissue images. We evaluate our few-shot colorectral tissue image generation by performing extensive qualitative, quantitative and subject specialist (pathologist) based evaluations. Specifically, in specialist-based evaluation, pathologists could differentiate between our XM-GAN generated tissue images and real images only 55% time. Moreover, we utilize these generated images as data augmentation to address the few-shot tissue image classification task, achieving a gain of 4.4% in terms of mean accuracy over the vanilla few-shot classifier. Code: \url{https://github.com/VIROBO-15/XM-GAN}
Geolocating precise locations from images presents a challenging problem in computer vision and information retrieval.Traditional methods typically employ either classification, which dividing the Earth surface into grid cells and classifying images accordingly, or retrieval, which identifying locations by matching images with a database of image-location pairs. However, classification-based approaches are limited by the cell size and cannot yield precise predictions, while retrieval-based systems usually suffer from poor search quality and inadequate coverage of the global landscape at varied scale and aggregation levels. To overcome these drawbacks, we present Img2Loc, a novel system that redefines image geolocalization as a text generation task. This is achieved using cutting-edge large multi-modality models like GPT4V or LLaVA with retrieval augmented generation. Img2Loc first employs CLIP-based representations to generate an image-based coordinate query database. It then uniquely combines query results with images itself, forming elaborate prompts customized for LMMs. When tested on benchmark datasets such as Im2GPS3k and YFCC4k, Img2Loc not only surpasses the performance of previous state-of-the-art models but does so without any model training.
We present a holistic approach for high resolution image classification that won second place in the ICCV/CVPPA2023 Deep Nutrient Deficiency Challenge. The approach consists of a full pipeline of: 1) data distribution analysis to check potential domain shift, 2) backbone selection for a strong baseline model that scales up for high resolution input, 3) transfer learning that utilizes published pretrained models and continuous fine-tuning on small sub-datasets, 4) data augmentation for the diversity of training data and to prevent overfitting, 5) test-time augmentation to improve the prediction's robustness, and 6) "data soups" that conducts cross-fold model prediction average for smoothened final test results.
There are a thousand ways to caption an image. Contrastive Language Pretraining (CLIP) on the other hand, works by mapping an image and its caption to a single vector -- limiting how well CLIP-like models can represent the diverse ways to describe an image. In this work, we introduce Llip, Latent Language Image Pretraining, which models the diversity of captions that could match an image. Llip's vision encoder outputs a set of visual features that are mixed into a final representation by conditioning on information derived from the text. We show that Llip outperforms non-contextualized baselines like CLIP and SigLIP on a variety of tasks even with large-scale encoders. Llip improves zero-shot classification by an average of 2.9% zero-shot classification benchmarks with a ViT-G/14 encoder. Specifically, Llip attains a zero-shot top-1 accuracy of 83.5% on ImageNet outperforming a similarly sized CLIP by 1.4%. We also demonstrate improvement on zero-shot retrieval on MS-COCO by 6.0%. We provide a comprehensive analysis of the components introduced by the method and demonstrate that Llip leads to richer visual representations.
Superpixels are higher-order perceptual groups of pixels in an image, often carrying much more information than the raw pixels. There is an inherent relational structure to the relationship among different superpixels of an image such as adjacent superpixels are neighbours of each other. Our interest here is to treat these relative positions of various superpixels as relational information of an image. This relational information can convey higher-order spatial information about the image, such as the relationship between superpixels representing two eyes in an image of a cat. That is, two eyes are placed adjacent to each other in a straight line or the mouth is below the nose. Our motive in this paper is to assist computer vision models, specifically those based on Deep Neural Networks (DNNs), by incorporating this higher-order information from superpixels. We construct a hybrid model that leverages (a) Convolutional Neural Network (CNN) to deal with spatial information in an image and (b) Graph Neural Network (GNN) to deal with relational superpixel information in the image. The proposed model is learned using a generic hybrid loss function. Our experiments are extensive, and we evaluate the predictive performance of our proposed hybrid vision model on seven different image classification datasets from a variety of domains such as digit and object recognition, biometrics, medical imaging. The results demonstrate that the relational superpixel information processed by a GNN can improve the performance of a standard CNN-based vision system.
Music Information Retrieval (MIR) has seen a recent surge in deep learning-based approaches, which often involve encoding symbolic music (i.e., music represented in terms of discrete note events) in an image-like or language like fashion. However, symbolic music is neither an image nor a sentence, and research in the symbolic domain lacks a comprehensive overview of the different available representations. In this paper, we investigate matrix (piano roll), sequence, and graph representations and their corresponding neural architectures, in combination with symbolic scores and performances on three piece-level classification tasks. We also introduce a novel graph representation for symbolic performances and explore the capability of graph representations in global classification tasks. Our systematic evaluation shows advantages and limitations of each input representation. Our results suggest that the graph representation, as the newest and least explored among the three approaches, exhibits promising performance, while being more light-weight in training.
Hate speech, quite common in the age of social media, at times harmless but can also cause mental trauma to someone or even riots in communities. Image of a religious symbol with derogatory comment or video of a man abusing a particular community, all become hate speech with its every modality (such as text, image, and audio) contributing towards it. Models based on a particular modality of hate speech post on social media are not useful, rather, we need models like multi-modal fusion models that consider both image and text while classifying hate speech. Text-image fusion models are heavily parameterized, hence we propose a quaternion neural network-based model having additional fusion components for each pair of modalities. The model is tested on the MMHS150K twitter dataset for hate speech classification. The model shows an almost 75% reduction in parameters and also benefits us in terms of storage space and training time while being at par in terms of performance as compared to its real counterpart.
The approaches for analyzing the polarimetric scattering matrix of polarimetric synthetic aperture radar (PolSAR) data have always been the focus of PolSAR image classification. Generally, the polarization coherent matrix and the covariance matrix obtained by the polarimetric scattering matrix only show a limited number of polarimetric information. In order to solve this problem, we propose a sparse scattering coding way to deal with polarimetric scattering matrix and obtain a close complete feature. This encoding mode can also maintain polarimetric information of scattering matrix completely. At the same time, in view of this encoding way, we design a corresponding classification algorithm based on convolution network to combine this feature. Based on sparse scattering coding and convolution neural network, the polarimetric convolutional network is proposed to classify PolSAR images by making full use of polarimetric information. We perform the experiments on the PolSAR images acquired by AIRSAR and RADARSAT-2 to verify the proposed method. The experimental results demonstrate that the proposed method get better results and has huge potential for PolSAR data classification. Source code for sparse scattering coding is available at https://github.com/liuxuvip/Polarimetric-Scattering-Coding.
Data augmentation is a crucial regularization technique for deep neural networks, particularly in medical image classification. Mainstream data augmentation (DA) methods are usually applied at the image level. Due to the specificity and diversity of medical imaging, expertise is often required to design effective DA strategies, and improper augmentation operations can degrade model performance. Although automatic augmentation methods exist, they are computationally intensive. Semantic data augmentation can implemented by translating features in feature space. However, over-translation may violate the image label. To address these issues, we propose \emph{Bayesian Random Semantic Data Augmentation} (BSDA), a computationally efficient and handcraft-free feature-level DA method. BSDA uses variational Bayesian to estimate the distribution of the augmentable magnitudes, and then a sample from this distribution is added to the original features to perform semantic data augmentation. We performed experiments on nine 2D and five 3D medical image datasets. Experimental results show that BSDA outperforms current DA methods. Additionally, BSDA can be easily assembled into CNNs or Transformers as a plug-and-play module, improving the network's performance. The code is available online at \url{https://github.com/YaoyaoZhu19/BSDA}.
Annotating words in a historical document image archive for word image recognition purpose demands time and skilled human resource (like historians, paleographers). In a real-life scenario, obtaining sample images for all possible words is also not feasible. However, Zero-shot learning methods could aptly be used to recognize unseen/out-of-lexicon words in such historical document images. Based on previous state-of-the-art method for zero-shot word recognition Pho(SC)Net, we propose a hybrid model based on the CTC framework (Pho(SC)-CTC) that takes advantage of the rich features learned by Pho(SC)Net followed by a connectionist temporal classification (CTC) framework to perform the final classification. Encouraging results were obtained on two publicly available historical document datasets and one synthetic handwritten dataset, which justifies the efficacy of Pho(SC)-CTC and Pho(SC)Net.
Interpreting and explaining the behavior of deep neural networks is critical for many tasks. Explainable AI provides a way to address this challenge, mostly by providing per-pixel relevance to the decision. Yet, interpreting such explanations may require expert knowledge. Some recent attempts toward interpretability adopt a concept-based framework, giving a higher-level relationship between some concepts and model decisions. This paper proposes Bottleneck Concept Learner (BotCL), which represents an image solely by the presence/absence of concepts learned through training over the target task without explicit supervision over the concepts. It uses self-supervision and tailored regularizers so that learned concepts can be human-understandable. Using some image classification tasks as our testbed, we demonstrate BotCL's potential to rebuild neural networks for better interpretability. Code is available at https://github.com/wbw520/BotCL and a simple demo is available at https://botcl.liangzhili.com/.