Fairness has been a critical issue that affects the adoption of deep learning models in real practice. To improve model fairness, many existing methods have been proposed and evaluated to be effective in their own contexts. However, there is still no systematic evaluation among them for a comprehensive comparison under the same context, which makes it hard to understand the performance distinction among them, hindering the research progress and practical adoption of them. To fill this gap, this paper endeavours to conduct the first large-scale empirical study to comprehensively compare the performance of existing state-of-the-art fairness improving techniques. Specifically, we target the widely-used application scenario of image classification, and utilized three different datasets and five commonly-used performance metrics to assess in total 13 methods from diverse categories. Our findings reveal substantial variations in the performance of each method across different datasets and sensitive attributes, indicating over-fitting on specific datasets by many existing methods. Furthermore, different fairness evaluation metrics, due to their distinct focuses, yield significantly different assessment results. Overall, we observe that pre-processing methods and in-processing methods outperform post-processing methods, with pre-processing methods exhibiting the best performance. Our empirical study offers comprehensive recommendations for enhancing fairness in deep learning models. We approach the problem from multiple dimensions, aiming to provide a uniform evaluation platform and inspire researchers to explore more effective fairness solutions via a set of implications.
Osteoporosis is a widespread and chronic metabolic bone disease that often remains undiagnosed and untreated due to limited access to bone mineral density (BMD) tests like Dual-energy X-ray absorptiometry (DXA). In response to this challenge, current advancements are pivoting towards detecting osteoporosis by examining alternative indicators from peripheral bone areas, with the goal of increasing screening rates without added expenses or time. In this paper, we present a method to predict osteoporosis using hand and wrist X-ray images, which are both widely accessible and affordable, though their link to DXA-based data is not thoroughly explored. We employ a sophisticated image segmentation model that utilizes a mixture of probabilistic U-Net decoders, specifically designed to capture predictive uncertainty in the segmentation of the ulna, radius, and metacarpal bones. This model is formulated as an optimal transport (OT) problem, enabling it to handle the inherent uncertainties in image segmentation more effectively. Further, we adopt a self-supervised learning (SSL) approach to extract meaningful representations without the need for explicit labels, and move on to classify osteoporosis in a supervised manner. Our method is evaluated on a dataset with 192 individuals, cross-referencing their verified osteoporosis conditions against the standard DXA test. With a notable classification score, this integration of uncertainty-aware segmentation and self-supervised learning represents a pioneering effort in leveraging vision-based techniques for the early detection of osteoporosis from peripheral skeletal sites.
The rapid advancement of generative AI models capable of creating realistic media has led to a need for classifiers that can accurately distinguish between genuine and artificially-generated images. A significant challenge for these classifiers emerges when they encounter images from generative models that are not represented in their training data, usually resulting in diminished performance. A typical approach is to periodically update the classifier's training data with images from the new generative models then retrain the classifier on the updated dataset. However, in some real-life scenarios, storage, computational, or privacy constraints render this approach impractical. Additionally, models used in security applications may be required to rapidly adapt. In these circumstances, continual learning provides a promising alternative, as the classifier can be updated without retraining on the entire dataset. In this paper, we introduce a new dataset called CLOFAI (Continual Learning On Fake and Authentic Images), which takes the form of a domain-incremental image classification problem. Moreover, we showcase the applicability of this dataset as a benchmark for evaluating continual learning methodologies. In doing this, we set a baseline on our novel dataset using three foundational continual learning methods -- EWC, GEM, and Experience Replay -- and find that EWC performs poorly, while GEM and Experience Replay show promise, performing significantly better than a Naive baseline. The dataset and code to run the experiments can be accessed from the following GitHub repository: https://github.com/Will-Doherty/CLOFAI.
Modern Artificial Intelligence (AI) increasingly relies on multi-agent architectures that blend visual and language understanding. Yet, a pressing challenge remains: How can we trust these agents especially in zero-shot settings with no fine-tuning? We introduce a novel modular Agentic AI visual classification framework that integrates generalist multimodal agents with a non-visual reasoning orchestrator and a Retrieval-Augmented Generation (RAG) module. Applied to apple leaf disease diagnosis, we benchmark three configurations: (I) zero-shot with confidence-based orchestration, (II) fine-tuned agents with improved performance, and (III) trust-calibrated orchestration enhanced by CLIP-based image retrieval and re-evaluation loops. Using confidence calibration metrics (ECE, OCR, CCC), the orchestrator modulates trust across agents. Our results demonstrate a 77.94\% accuracy improvement in the zero-shot setting using trust-aware orchestration and RAG, achieving 85.63\% overall. GPT-4o showed better calibration, while Qwen-2.5-VL displayed overconfidence. Furthermore, image-RAG grounded predictions with visually similar cases, enabling correction of agent overconfidence via iterative re-evaluation. The proposed system separates perception (vision agents) from meta-reasoning (orchestrator), enabling scalable and interpretable multi-agent AI. This blueprint is extensible to diagnostics, biology, and other trust-critical domains. All models, prompts, results, and system components including the complete software source code are openly released to support reproducibility, transparency, and community benchmarking at Github: https://github.com/Applied-AI-Research-Lab/Orchestrator-Agent-Trust
비전 트랜스포머(ViT)는 주로 셀프 어텐션 메커니즘 덕분에 다양한 컴퓨터 비전 작업에서 최첨단 성능을 달성했다. 그러나 토큰 수에 따라 제곱으로 증가하는 그 복잡도는 ViT의 효율성과 대규모 배포에 여전히 주요한 장애물이다. 토큰 병합은 중복된 토큰을 집약하여 이 비용을 줄인다. 하지만 기존 방법은 대개 단일 아키텍처 내에서 평가되어, 그 효과가 병합 메커니즘 자체에서 비롯되는지 아니면 함께 사용된 특정 디코더에서 비롯되는지가 불분명하다. 우리는 ViT 기반 네트워크 초기에 삽입되는 단일 모듈인 그래프 유도 토큰 병합(G2TM)을 원래의 Segmenter 설정 너머로 확장한다. G2TM을 세 가지 의미론적 분할 프레임워크(Segmenter, SETR, EoMT)와 세 가지 디코더 계열(선형, 트랜스포머 기반, 합성곱 기반)은 물론 표준 ViT 이미지 분류에서 평가한다. 결과는 G2TM의 거동과 정확도-효율 간 절충이 주어진 백본 크기에서 시험한 모든 아키텍처에 걸쳐 일관됨을 보이며, 그 효과가 디코더가 아니라 인코더의 속성임을 시사한다. G2TM은 이미지 분류에도 잘 일반화되어 의미론적 분할에 비해 정확도 저하가 더 작다. 또한 ADE20K 데이터셋의 분할 모델에서 GFLOPs를 일관되게 22-47% 줄이고 처리량을 최대 74% 높이는 G2TM의 최적 하이퍼파라미터가 디코더 선택보다는 주로 백본의 사전학습 방식과 대상 데이터셋에 의존함을 확인한다.
Crowdsourcing has become widely used in supervised scenarios where training sets are scarce and difficult to obtain. Most crowdsourcing models in the literature assume labelers can provide answers to full questions. In classification contexts, full questions require a labeler to discern among all possible classes. Unfortunately, discernment is not always easy in realistic scenarios. Labelers may not be experts in differentiating all classes. In this work, we provide a full probabilistic model for a shorter type of queries. Our shorter queries only require "yes" or "no" responses. Our model estimates a joint posterior distribution of matrices related to labelers' confusions and the posterior probability of the class of every object. We developed an approximate inference approach, using Monte Carlo Sampling and Black Box Variational Inference, which provides the derivation of the necessary gradients. We built two realistic crowdsourcing scenarios to test our model. The first scenario queries for irregular astronomical time-series. The second scenario relies on the image classification of animals. We achieved results that are comparable with those of full query crowdsourcing. Furthermore, we show that modeling labelers' failures plays an important role in estimating true classes. Finally, we provide the community with two real datasets obtained from our crowdsourcing experiments. All our code is publicly available.
In this paper, we show how convolutional neural networks (CNN) can be used in regression and classification learning problems of noisy and non-noisy functional data. The main idea is to transform the functional data into a 28 by 28 image. We use a specific but typical architecture of a convolutional neural network to perform all the regression exercises of parameter estimation and functional form classification. First, we use some functional case studies of functional data with and without random noise to showcase the strength of the new method. In particular, we use it to estimate exponential growth and decay rates, the bandwidths of sine and cosine functions, and the magnitudes and widths of curve peaks. We also use it to classify the monotonicity and curvatures of functional data, algebraic versus exponential growth, and the number of peaks of functional data. Second, we apply the same convolutional neural networks to Lyapunov exponent estimation in noisy and non-noisy chaotic data, in estimating rates of disease transmission from epidemic curves, and in detecting the similarity of drug dissolution profiles. Finally, we apply the method to real-life data to detect Parkinson's disease patients in a classification problem. The method, although simple, shows high accuracy and is promising for future use in engineering and medical applications.
Few-shot anomaly detection methods can effectively address data collecting difficulty in industrial scenarios. Compared to 2D few-shot anomaly detection (2D-FSAD), 3D few-shot anomaly detection (3D-FSAD) is still an unexplored but essential task. In this paper, we propose CLIP3D-AD, an efficient 3D-FSAD method extended on CLIP. We successfully transfer strong generalization ability of CLIP into 3D-FSAD. Specifically, we synthesize anomalous images on given normal images as sample pairs to adapt CLIP for 3D anomaly classification and segmentation. For classification, we introduce an image adapter and a text adapter to fine-tune global visual features and text features. Meanwhile, we propose a coarse-to-fine decoder to fuse and facilitate intermediate multi-layer visual representations of CLIP. To benefit from geometry information of point cloud and eliminate modality and data discrepancy when processed by CLIP, we project and render point cloud to multi-view normal and anomalous images. Then we design multi-view fusion module to fuse features of multi-view images extracted by CLIP which are used to facilitate visual representations for further enhancing vision-language correlation. Extensive experiments demonstrate that our method has a competitive performance of 3D few-shot anomaly classification and segmentation on MVTec-3D AD dataset.
Semi-supervised learning methods have shown promising results in solving many practical problems when only a few labels are available. The existing methods assume that the class distributions of labeled and unlabeled data are equal; however, their performances are significantly degraded in class distribution mismatch scenarios where out-of-distribution (OOD) data exist in the unlabeled data. Previous safe semi-supervised learning studies have addressed this problem by making OOD data less likely to affect training based on labeled data. However, even if the studies effectively filter out the unnecessary OOD data, they can lose the basic information that all data share regardless of class. To this end, we propose to apply a self-supervised contrastive learning approach to fully exploit a large amount of unlabeled data. We also propose a contrastive loss function with coefficient schedule to aggregate as an anchor the labeled negative examples of the same class into positive examples. To evaluate the performance of the proposed method, we conduct experiments on image classification datasets - CIFAR-10, CIFAR-100, Tiny ImageNet, and CIFAR-100+Tiny ImageNet - under various mismatch ratios. The results show that self-supervised contrastive learning significantly improves classification accuracy. Moreover, aggregating the in-distribution examples produces better representation and consequently further improves classification accuracy.
Deep learning architectures exhibit a critical drop of performance due to catastrophic forgetting when they are required to incrementally learn new tasks. Contemporary incremental learning frameworks focus on image classification and object detection while in this work we formally introduce the incremental learning problem for semantic segmentation in which a pixel-wise labeling is considered. To tackle this task we propose to distill the knowledge of the previous model to retain the information about previously learned classes, whilst updating the current model to learn the new ones. We propose various approaches working both on the output logits and on intermediate features. In opposition to some recent frameworks, we do not store any image from previously learned classes and only the last model is needed to preserve high accuracy on these classes. The experimental evaluation on the Pascal VOC2012 dataset shows the effectiveness of the proposed approaches.
Much effort has been devoted to understanding the decisions of deep neural networks in recent years. A number of model-aware saliency methods were proposed to explain individual classification decisions by creating saliency maps. However, they are not applicable when the parameters and the gradients of the underlying models are unavailable. Recently, model-agnostic methods have also received attention. As one of them, \textit{Prediction Difference Analysis} (PDA), a probabilistic sound methodology, was proposed. In this work, we first show that PDA can suffer from saturated classifiers. The saturation phenomenon of classifiers exists widely in current neural network-based classifiers. To explain the decisions of saturated classifiers better, we further propose Contextual PDA, which runs hundreds of times faster than PDA. The experiments show the superiority of our method by explaining image classifications of the state-of-the-art deep convolutional neural networks.
Data augmentation can effectively resolve a scarcity of images when training machine-learning algorithms. It can make them more robust to unseen images. We present a lesion conditional Generative Adversarial Network LcGAN to generate synthetic Computed Tomography (CT) images for data augmentation. A lesion conditional image (segmented mask) is an input to both the generator and the discriminator of the LcGAN during training. The trained model generates contextual CT images based on input masks. We quantify the quality of the images by using a fully convolutional network (FCN) score and blurriness. We also train another classification network to select better synthetic images. These synthetic CT images are then augmented to our hemorrhagic lesion segmentation network. By applying this augmentation method on 2.5%, 10% and 25% of original data, segmentation improved by 12.8%, 6% and 1.6% respectively.
Model explanation techniques play a critical role in understanding the source of a model's performance and making its decisions transparent. Here we investigate if explanation techniques can also be used as a mechanism for scientific discovery. We make three contributions: first, we propose a framework to convert predictions from explanation techniques to a mechanism of discovery. Second, we show how generative models in combination with black-box predictors can be used to generate hypotheses (without human priors) that can be critically examined. Third, with these techniques we study classification models for retinal images predicting Diabetic Macular Edema (DME), where recent work showed that a CNN trained on these images is likely learning novel features in the image. We demonstrate that the proposed framework is able to explain the underlying scientific mechanism, thus bridging the gap between the model's performance and human understanding.
Vision Transformer (ViT) suffers from data scarcity in semi-supervised learning (SSL). To alleviate this issue, inspired by masked autoencoder (MAE), which is a data-efficient self-supervised learner, we propose Semi-MAE, a pure ViT-based SSL framework consisting of a parallel MAE branch to assist the visual representation learning and make the pseudo labels more accurate. The MAE branch is designed as an asymmetric architecture consisting of a lightweight decoder and a shared-weights encoder. We feed the weakly-augmented unlabeled data with a high masking ratio to the MAE branch and reconstruct the missing pixels. Semi-MAE achieves 75.9% top-1 accuracy on ImageNet with 10% labels, surpassing prior state-of-the-art in semi-supervised image classification. In addition, extensive experiments demonstrate that Semi-MAE can be readily used for other ViT models and masked image modeling methods.
Modern neural networks have revolutionized the fields of computer vision (CV) and Natural Language Processing (NLP). They are widely used for solving complex CV tasks and NLP tasks such as image classification, image generation, and machine translation. Most state-of-the-art neural networks are over-parameterized and require a high computational cost. One straightforward solution is to replace the layers of the networks with their low-rank tensor approximations using different tensor decomposition methods. This paper reviews six tensor decomposition methods and illustrates their ability to compress model parameters of convolutional neural networks (CNNs), recurrent neural networks (RNNs) and Transformers. The accuracy of some compressed models can be higher than the original versions. Evaluations indicate that tensor decompositions can achieve significant reductions in model size, run-time and energy consumption, and are well suited for implementing neural networks on edge devices.
Interpretable part-prototype models are computer vision models that are explainable by design. The models learn prototypical parts and recognise these components in an image, thereby combining classification and explanation. Despite the recent attention for intrinsically interpretable models, there is no comprehensive overview on evaluating the explanation quality of interpretable part-prototype models. Based on the Co-12 properties for explanation quality as introduced in arXiv:2201.08164 (e.g., correctness, completeness, compactness), we review existing work that evaluates part-prototype models, reveal research gaps and outline future approaches for evaluation of the explanation quality of part-prototype models. This paper, therefore, contributes to the progression and maturity of this relatively new research field on interpretable part-prototype models. We additionally provide a ``Co-12 cheat sheet'' that acts as a concise summary of our findings on evaluating part-prototype models.
In computer vision, models must be able to adapt to changes in image resolution to effectively carry out tasks such as image segmentation; This is known as scale-equivariance. Recent works have made progress in developing scale-equivariant convolutional neural networks, e.g., through weight-sharing and kernel resizing. However, these networks are not truly scale-equivariant in practice. Specifically, they do not consider anti-aliasing as they formulate the down-scaling operation in the continuous domain. To address this shortcoming, we directly formulate down-scaling in the discrete domain with consideration of anti-aliasing. We then propose a novel architecture based on Fourier layers to achieve truly scale-equivariant deep nets, i.e., absolute zero equivariance-error. Following prior works, we test this model on MNIST-scale and STL-10 datasets. Our proposed model achieves competitive classification performance while maintaining zero equivariance-error.
Latent image representations arising from vision-language models have proved immensely useful for a variety of downstream tasks. However, their utility is limited by their entanglement with respect to different visual attributes. For instance, recent work has shown that CLIP image representations are often biased toward specific visual properties (such as objects or actions) in an unpredictable manner. In this paper, we propose to separate representations of the different visual modalities in CLIP's joint vision-language space by leveraging the association between parts of speech and specific visual modes of variation (e.g. nouns relate to objects, adjectives describe appearance). This is achieved by formulating an appropriate component analysis model that learns subspaces capturing variability corresponding to a specific part of speech, while jointly minimising variability to the rest. Such a subspace yields disentangled representations of the different visual properties of an image or text in closed form while respecting the underlying geometry of the manifold on which the representations lie. What's more, we show the proposed model additionally facilitates learning subspaces corresponding to specific visual appearances (e.g. artists' painting styles), which enables the selective removal of entire visual themes from CLIP-based text-to-image synthesis. We validate the model both qualitatively, by visualising the subspace projections with a text-to-image model and by preventing the imitation of artists' styles, and quantitatively, through class invariance metrics and improvements to baseline zero-shot classification.
Contrastive pretraining of image-text foundation models, such as CLIP, demonstrated excellent zero-shot performance and improved robustness on a wide range of downstream tasks. However, these models utilize large transformer-based encoders with significant memory and latency overhead which pose challenges for deployment on mobile devices. In this work, we introduce MobileCLIP -- a new family of efficient image-text models optimized for runtime performance along with a novel and efficient training approach, namely multi-modal reinforced training. The proposed training approach leverages knowledge transfer from an image captioning model and an ensemble of strong CLIP encoders to improve the accuracy of efficient models. Our approach avoids train-time compute overhead by storing the additional knowledge in a reinforced dataset. MobileCLIP sets a new state-of-the-art latency-accuracy tradeoff for zero-shot classification and retrieval tasks on several datasets. Our MobileCLIP-S2 variant is 2.3$\times$ faster while more accurate compared to previous best CLIP model based on ViT-B/16. We further demonstrate the effectiveness of our multi-modal reinforced training by training a CLIP model based on ViT-B/16 image backbone and achieving +2.9% average performance improvement on 38 evaluation benchmarks compared to the previous best. Moreover, we show that the proposed approach achieves 10$\times$-1000$\times$ improved learning efficiency when compared with non-reinforced CLIP training. Code and models are available at https://github.com/apple/ml-mobileclip .
Deep learning models (DLMs) frequently achieve accurate segmentation and classification of tumors from medical images. However, DLMs lacking feedback on their image segmentation mechanisms, such as Dice coefficients and confidence in their performance, face challenges when processing previously unseen images in real-world clinical settings. Uncertainty estimates to identify DLM predictions at the cellular or single-pixel level that require clinician review can enhance trust. However, their deployment requires significant computational resources. This study reports two DLMs, one trained from scratch and another based on transfer learning, with Monte Carlo dropout or Bayes-by-backprop uncertainty estimations to segment lesions from the publicly available The International Skin Imaging Collaboration-19 dermoscopy image database with cancerous lesions. A novel approach to compute pixel-by-pixel uncertainty estimations of DLM segmentation performance in multiple clinical regions from a single dermoscopy image with corresponding Dice scores is reported for the first time. Image-level uncertainty maps demonstrated correspondence between imperfect DLM segmentation and high uncertainty levels in specific skin tissue regions, with or without lesions. Four new linear regression models that can predict the Dice performance of DLM segmentation using constants and uncertainty measures, either individually or in combination from lesions, tissue structures, and non-tissue pixel regions critical for clinical diagnosis and prognostication in skin images (Spearman's correlation, p < 0.05), are reported for the first time for low-compute uncertainty estimation workflows.