As performance gains through scaling data and/or model size experience diminishing returns, it is becoming increasingly popular to turn to ensembling, where the predictions of multiple models are combined to improve accuracy. In this paper, we provide a detailed analysis of how the disagreement and the polarization (a notion we introduce and define in this paper) among classifiers relate to the performance gain achieved by aggregating individual classifiers, for majority vote strategies in classification tasks. We address these questions in the following ways. (1) An upper bound for polarization is derived, and we propose what we call a neural polarization law: most interpolating neural network models are 4/3-polarized. Our empirical results not only support this conjecture but also show that polarization is nearly constant for a dataset, regardless of hyperparameters or architectures of classifiers. (2) The error of the majority vote classifier is considered under restricted entropy conditions, and we present a tight upper bound that indicates that the disagreement is linearly correlated with the target, and that the slope is linear in the polarization. (3) We prove results for the asymptotic behavior of the disagreement in terms of the number of classifiers, which we show can help in predicting the performance for a larger number of classifiers from that of a smaller number. Our theories and claims are supported by empirical results on several image classification tasks with various types of neural networks.
This study focuses on incremental learning for image classification, exploring how to reduce catastrophic forgetting of all learned knowledge when access to old data is restricted due to memory or privacy constraints. The challenge of incremental learning lies in achieving an optimal balance between plasticity, the ability to learn new knowledge, and stability, the ability to retain old knowledge. Based on whether the task identifier (task-ID) of an image can be obtained during the test stage, incremental learning for image classifcation is divided into two main paradigms, which are task incremental learning (TIL) and class incremental learning (CIL). The TIL paradigm has access to the task-ID, allowing it to use multiple task-specific classification heads selected based on the task-ID. Consequently, in CIL, where the task-ID is unavailable, TIL methods must predict the task-ID to extend their application to the CIL paradigm. Our previous method for TIL adds task-specific batch normalization and classification heads incrementally. This work extends the method by predicting task-ID through an "unknown" class added to each classification head. The head with the lowest "unknown" probability is selected, enabling task-ID prediction and making the method applicable to CIL. The task-specific batch normalization (BN) modules effectively adjust the distribution of output feature maps across different tasks, enhancing the model's plasticity.Moreover, since BN has much fewer parameters compared to convolutional kernels, by only modifying the BN layers as new tasks arrive, the model can effectively manage parameter growth while ensuring stability across tasks. The innovation of this study lies in the first-time introduction of task-specific BN into CIL and verifying the feasibility of extending TIL methods to CIL through task-ID prediction with state-of-the-art performance on multiple datasets.
This study presents an innovative dynamic weighting knowledge distillation (KD) framework tailored for efficient Earth observation (EO) image classification (IC) in resource-constrained settings. Utilizing EfficientViT and MobileViT as teacher models, this framework enables lightweight student models, particularly ResNet8 and ResNet16, to surpass 90% in accuracy, precision, and recall, adhering to the stringent confidence thresholds necessary for reliable classification tasks. Unlike conventional KD methods that rely on static weight distribution, our adaptive weighting mechanism responds to each teacher model's confidence, allowing student models to prioritize more credible sources of knowledge dynamically. Remarkably, ResNet8 delivers substantial efficiency gains, achieving a 97.5% reduction in parameters, a 96.7% decrease in FLOPs, an 86.2% cut in power consumption, and a 63.5% increase in inference speed over MobileViT. This significant optimization of complexity and resource demands establishes ResNet8 as an optimal candidate for EO tasks, combining robust performance with feasibility in deployment. The confidence-based, adaptable KD approach underscores the potential of dynamic distillation strategies to yield high-performing, resource-efficient models tailored for satellite-based EO applications. The reproducible code is accessible on our GitHub repository.
Countries in South Asia experience many catastrophic flooding events regularly. Through image classification, it is possible to expedite search and rescue initiatives by classifying flood zones, including houses and humans. We create a new dataset collecting aerial imagery of flooding events across South Asian countries. For the classification, we propose a fine-tuned Compact Convolutional Transformer (CCT) based approach and some other cutting-edge transformer-based and Convolutional Neural Network-based architectures (CNN). We also implement the YOLOv8 object detection model and detect houses and humans within the imagery of our proposed dataset, and then compare the performance with our classification-based approach. Since the countries in South Asia have similar topography, housing structure, the color of flood water, and vegetation, this work can be more applicable to such a region as opposed to the rest of the world. The images are divided evenly into four classes: 'flood', 'flood with domicile', 'flood with humans', and 'no flood'. After experimenting with our proposed dataset on our fine-tuned CCT model, which has a comparatively lower number of weight parameters than many other transformer-based architectures designed for computer vision, it exhibits an accuracy and macro average precision of 98.62% and 98.50%. The other transformer-based architectures that we implement are the Vision Transformer (ViT), Swin Transformer, and External Attention Transformer (EANet), which give an accuracy of 88.66%, 84.74%, and 66.56% respectively. We also implement DCECNN (Deep Custom Ensembled Convolutional Neural Network), which is a custom ensemble model that we create by combining MobileNet, InceptionV3, and EfficientNetB0, and we obtain an accuracy of 98.78%. The architectures we implement are fine-tuned to achieve optimal performance on our dataset.
The COVID19 pandemic has had a detrimental impact on the health and welfare of the worlds population. An important strategy in the fight against COVID19 is the effective screening of infected patients, with one of the primary screening methods involving radiological imaging with the use of chest Xrays. This is why this study introduces a customized convolutional neural network (CCNN) for medical image classification. This study used a dataset of 6432 images named Chest Xray (COVID19 and Pneumonia), and images were preprocessed using techniques, including resizing, normalizing, and augmentation, to improve model training and performance. The proposed CCNN was compared with a convolutional neural network (CNN) and other models that used the same dataset. This research found that the Convolutional Neural Network (CCNN) achieved 95.62% validation accuracy and 0.1270 validation loss. This outperformed earlier models and studies using the same dataset. This result indicates that our models learn effectively from training data and adapt efficiently to new, unseen data. In essence, the current CCNN model achieves better medical image classification performance, which is why this CCNN model efficiently classifies medical images. Future research may extend the models application to other medical imaging datasets and develop realtime offline medical image classification websites or apps.
We propose Few-Class Arena (FCA), as a unified benchmark with focus on testing efficient image classification models for few classes. A wide variety of benchmark datasets with many classes (80-1000) have been created to assist Computer Vision architectural evolution. An increasing number of vision models are evaluated with these many-class datasets. However, real-world applications often involve substantially fewer classes of interest (2-10). This gap between many and few classes makes it difficult to predict performance of the few-class applications using models trained on the available many-class datasets. To date, little has been offered to evaluate models in this Few-Class Regime. We conduct a systematic evaluation of the ResNet family trained on ImageNet subsets from 2 to 1000 classes, and test a wide spectrum of Convolutional Neural Networks and Transformer architectures over ten datasets by using our newly proposed FCA tool. Furthermore, to aid an up-front assessment of dataset difficulty and a more efficient selection of models, we incorporate a difficulty measure as a function of class similarity. FCA offers a new tool for efficient machine learning in the Few-Class Regime, with goals ranging from a new efficient class similarity proposal, to lightweight model architecture design, to a new scaling law. FCA is user-friendly and can be easily extended to new models and datasets, facilitating future research work. Our benchmark is available at https://github.com/fewclassarena/fca.
Detecting genetic aberrations is crucial in cancer diagnosis, typically through fluorescence in situ hybridization (FISH). However, existing FISH image classification methods face challenges due to signal variability, the need for costly manual annotations and fail to adequately address the intrinsic uncertainty. We introduce a novel approach that leverages synthetic images to eliminate the requirement for manual annotations and utilizes a joint contrastive and classification objective for training to account for inter-class variation effectively. We demonstrate the superior generalization capabilities and uncertainty calibration of our method, which is trained on synthetic data, by testing it on a manually annotated dataset of real-world FISH images. Our model offers superior calibration in terms of classification accuracy and uncertainty quantification with a classification accuracy of 96.7% among the 50% most certain cases. The presented end-to-end method reduces the demands on personnel and time and improves the diagnostic workflow due to its accuracy and adaptability. All code and data is publicly accessible at: https://github.com/SimonBon/FISHing
Low-resource domains, characterized by scarce data and annotations, present significant challenges for language and visual understanding tasks, with the latter much under-explored in the literature. Recent advancements in Vision-Language Models (VLM) have shown promising results in high-resource domains but fall short in low-resource concepts that are under-represented (e.g. only a handful of images per category) in the pre-training set. We tackle the challenging task of zero-shot low-resource image classification from a novel perspective. By leveraging a retrieval-based strategy, we achieve this in a training-free fashion. Specifically, our method, named CoRE (Combination of Retrieval Enrichment), enriches the representation of both query images and class prototypes by retrieving relevant textual information from large web-crawled databases. This retrieval-based enrichment significantly boosts classification performance by incorporating the broader contextual information relevant to the specific class. We validate our method on a newly established benchmark covering diverse low-resource domains, including medical imaging, rare plants, and circuits. Our experiments demonstrate that CORE outperforms existing state-of-the-art methods that rely on synthetic data generation and model fine-tuning.
Objects, in the real world, rarely occur in isolation and exhibit typical arrangements governed by their independent utility, and their expected interaction with humans and other objects in the context. For example, a chair is expected near a table, and a computer is expected on top. Humans use this spatial context and relative placement as an important cue for visual recognition in case of ambiguities. Similar to human's, DNN's exploit contextual information from data to learn representations. Our research focuses on harnessing the contextual aspects of visual data to optimize data annotation and enhance the training of deep networks. Our contributions can be summarized as follows: (1) We introduce the notion of contextual diversity for active learning CDAL and show its applicability in three different visual tasks semantic segmentation, object detection and image classification, (2) We propose a data repair algorithm to curate contextually fair data to reduce model bias, enabling the model to detect objects out of their obvious context, (3) We propose Class-based annotation, where contextually relevant classes are selected that are complementary for model training under domain shift. Understanding the importance of well-curated data, we also emphasize the necessity of involving humans in the loop to achieve accurate annotations and to develop novel interaction strategies that allow humans to serve as fact-checkers. In line with this we are working on developing image retrieval system for wildlife camera trap images and reliable warning system for poor quality rural roads. For large-scale annotation, we are employing a strategic combination of human expertise and zero-shot models, while also integrating human input at various stages for continuous feedback.
Deep learning has excelled in medical image classification, but its clinical application is limited by poor interpretability. Capsule networks, known for encoding hierarchical relationships and spatial features, show potential in addressing this issue. Nevertheless, traditional capsule networks often underperform due to their shallow structures, and deeper variants lack hierarchical architectures, thereby compromising interpretability. This paper introduces a novel capsule network, ParseCaps, which utilizes the sparse axial attention routing and parse convolutional capsule layer to form a parse-tree-like structure, enhancing both depth and interpretability. Firstly, sparse axial attention routing optimizes connections between child and parent capsules, as well as emphasizes the weight distribution across instantiation parameters of parent capsules. Secondly, the parse convolutional capsule layer generates capsule predictions aligning with the parse tree. Finally, based on the loss design that is effective whether concept ground truth exists or not, ParseCaps advances interpretability by associating each dimension of the global capsule with a comprehensible concept, thereby facilitating clinician trust and understanding of the model's classification results. Experimental results on CE-MRI, PH$^2$, and Derm7pt datasets show that ParseCaps not only outperforms other capsule network variants in classification accuracy, redundancy reduction and robustness, but also provides interpretable explanations, regardless of the availability of concept labels.
Deep neural networks (DNNs) have recently achieved impressive success across a wide range of real-world vision and language processing tasks, spanning from image classification to many other downstream vision tasks, such as object detection, tracking, and segmentation. However, previous well-established DNNs, despite being able to maintain superior accuracy, have also been evolving to be deeper and wider and thus inevitably necessitate prohibitive computational resources for both training and inference. This trend further enlarges the computational gap between computation-intensive DNNs and resource-constrained embedded computing systems, making it challenging to deploy powerful DNNs upon real-world embedded computing systems towards ubiquitous embedded intelligence. To alleviate the above computational gap and enable ubiquitous embedded intelligence, we, in this survey, focus on discussing recent efficient deep learning infrastructures for embedded computing systems, spanning from training to inference, from manual to automated, from convolutional neural networks to transformers, from transformers to vision transformers, from vision models to large language models, from software to hardware, and from algorithms to applications. Specifically, we discuss recent efficient deep learning infrastructures for embedded computing systems from the lens of (1) efficient manual network design for embedded computing systems, (2) efficient automated network design for embedded computing systems, (3) efficient network compression for embedded computing systems, (4) efficient on-device learning for embedded computing systems, (5) efficient large language models for embedded computing systems, (6) efficient deep learning software and hardware for embedded computing systems, and (7) efficient intelligent applications for embedded computing systems.
In recent years, the diagnosis of gastrointestinal (GI) diseases has advanced greatly with the advent of high-tech video capsule endoscopy (VCE) technology, which allows for non-invasive observation of the digestive system. The MisaHub Capsule Vision Challenge encourages the development of vendor-independent artificial intelligence models that can autonomously classify GI anomalies from VCE images. This paper presents CNN architecture designed specifically for multiclass classification of ten gut pathologies, including angioectasia, bleeding, erosion, erythema, foreign bodies, lymphangiectasia, polyps, ulcers, and worms as well as their normal state.
Video capsule endoscopy has transformed gastrointestinal endoscopy (GIE) diagnostics by offering a non-invasive method for capturing detailed images of the gastrointestinal tract, enabling early disease detection. However, its potential is limited by the sheer volume of images generated during the imaging procedure, which can take anywhere from 6-8 hours and often produce up to 1 million images, necessitating automated analysis. Additionally, the variability of these images, combined with the need for expert annotations and the scarcity of large, high-quality labeled datasets, constrains the effectiveness of current medical image analysis models. To address this, we introduce a novel large GIE dataset, called EndoExtend24, created by merging ten existing public and private datasets, ensuring patient integrity across splits. EndoExtend24 includes over 226,000 labeled images, as well as dynamic class mappings, which allow unified training across datasets with differing labeling granularity, supporting up to 123 distinct pathological findings. Further, we propose to leverage domain adaptive pre-training of foundation models trained with self-supervision on generic image data, to adapt them to the task of GIE medical image diagnosis. Specifically, the EVA-02 model, which is based on the ViT architecture and trained on ImageNet-22k with masked image modeling (using EVA-CLIP as a MIM teacher), is pre-trained on the EndoExtend24 dataset to achieve domain adaptation, and finally trained on the Capsule Endoscopy 2024 Challenge dataset. Our model demonstrates robust performance, securing third place in the Capsule Endoscopy 2024 Challenge. We achieved a macro AUC of 0.762 and a balanced accuracy of 37.1% on the test set. These results emphasize the effectiveness of our domain-adaptive pre-training approach and the enriched EndoExtend24 dataset in advancing gastrointestinal endoscopy diagnostics.
Methods based on Contrastive Language-Image Pre-training (CLIP) are nowadays extensively used in support of vision-and-language tasks involving remote sensing data, such as cross-modal retrieval. The adaptation of CLIP to this specific domain has relied on model fine-tuning with the standard contrastive objective, using existing human-labeled image-caption datasets, or using synthetic data corresponding to image-caption pairs derived from other annotations over remote sensing images (e.g., object classes). The use of different pre-training mechanisms has received less attention, and only a few exceptions have considered multilingual inputs. This work proposes a novel vision-and-language model for the remote sensing domain, exploring the fine-tuning of a multilingual CLIP model and testing the use of a self-supervised method based on aligning local and global representations from individual input images, together with the standard CLIP objective. Model training relied on assembling pre-existing datasets of remote sensing images paired with English captions, followed by the use of automated machine translation into nine additional languages. We show that translated data is indeed helpful, e.g. improving performance also on English. Our resulting model, which we named Remote Sensing Multilingual CLIP (RS-M-CLIP), obtains state-of-the-art results in a variety of vision-and-language tasks, including cross-modal and multilingual image-text retrieval, or zero-shot image classification.
In many modern computer application problems, the classification of image data plays an important role. Among many different supervised machine learning models, convolutional neural networks (CNNs) and linear discriminant analysis (LDA) as well as sophisticated variants thereof are popular techniques. In this work, two different domain decomposed CNN models are experimentally compared for different image classification problems. Both models are loosely inspired by domain decomposition methods and in addition, combined with a transfer learning strategy. The resulting models show improved classification accuracies compared to the corresponding, composed global CNN model without transfer learning and besides, also help to speed up the training process. Moreover, a novel decomposed LDA strategy is proposed which also relies on a localization approach and which is combined with a small neural network model. In comparison with a global LDA applied to the entire input data, the presented decomposed LDA approach shows increased classification accuracies for the considered test problems.
Machine unlearning (MU) has gained significant attention as a means to remove specific data from trained models without requiring a full retraining process. While progress has been made in unimodal domains like text and image classification, unlearning in multimodal models remains relatively underexplored. In this work, we address the unique challenges of unlearning in CLIP, a prominent multimodal model that aligns visual and textual representations. We introduce CLIPErase, a novel approach that disentangles and selectively forgets both visual and textual associations, ensuring that unlearning does not compromise model performance. CLIPErase consists of three key modules: a Forgetting Module that disrupts the associations in the forget set, a Retention Module that preserves performance on the retain set, and a Consistency Module that maintains consistency with the original model. Extensive experiments on the CIFAR-100 and Flickr30K datasets across four CLIP downstream tasks demonstrate that CLIPErase effectively forgets designated associations in zero-shot tasks for multimodal samples, while preserving the model's performance on the retain set after unlearning.
Neural Architecture Search (NAS) methods autonomously discover high-accuracy neural network architectures, outperforming manually crafted ones. However, The NAS methods require high computational costs due to the high dimension search space and the need to train multiple candidate solutions. This paper introduces LCoDeepNEAT, an instantiation of Lamarckian genetic algorithms, which extends the foundational principles of the CoDeepNEAT framework. LCoDeepNEAT co-evolves CNN architectures and their respective final layer weights. The evaluation process of LCoDeepNEAT entails a single epoch of SGD, followed by the transference of the acquired final layer weights to the genetic representation of the network. In addition, it expedites the process of evolving by imposing restrictions on the architecture search space, specifically targeting architectures comprising just two fully connected layers for classification. Our method yields a notable improvement in the classification accuracy of candidate solutions throughout the evolutionary process, ranging from 2% to 5.6%. This outcome underscores the efficacy and effectiveness of integrating gradient information and evolving the last layer of candidate solutions within LCoDeepNEAT. LCoDeepNEAT is assessed across six standard image classification datasets and benchmarked against eight leading NAS methods. Results demonstrate LCoDeepNEAT's ability to swiftly discover competitive CNN architectures with fewer parameters, conserving computational resources, and achieving superior classification accuracy compared to other approaches.
Breast cancer histopathology image classification is critical for early detection and improved patient outcomes. 1 This study introduces a novel approach leveraging EfficientNetV2 models, to improve feature extraction and focus on relevant tissue regions. The proposed models were evaluated on the BreakHis dataset across multiple magnification scales (40X, 100X, 200X, and 400X). 2 Among them, the EfficientNetV2-XL with CBAM achieved outstanding performance, reaching a peak accuracy of 99.01 percent and an F1-score of 98.31 percent at 400X magnification, outperforming state-of-the-art methods. 3 By integrating Contrast Limited Adaptive Histogram Equalization (CLAHE) for preprocessing and optimizing computational efficiency, this method demonstrates its suitability for real-time clinical deployment. 3 The results underscore the potential of attention-enhanced scalable architectures in advancing diagnostic precision for breast cancer detection.
Pre-trained vision-language models (VLMs) like CLIP have demonstrated impressive zero-shot performance on a wide range of downstream computer vision tasks. However, there still exists a considerable performance gap between these models and a supervised deep model trained on a downstream dataset. To bridge this gap, we propose a novel active learning (AL) framework that enhances the zero-shot classification performance of VLMs by selecting only a few informative samples from the unlabeled data for annotation during training. To achieve this, our approach first calibrates the predicted entropy of VLMs and then utilizes a combination of self-uncertainty and neighbor-aware uncertainty to calculate a reliable uncertainty measure for active sample selection. Our extensive experiments show that the proposed approach outperforms existing AL approaches on several image classification datasets, and significantly enhances the zero-shot performance of VLMs.
Recently, Vision Transformers (ViTs) have achieved unprecedented effectiveness in the general domain of image classification. Nonetheless, these models remain underexplored in the field of deepfake detection, given their lower performance as compared to Convolution Neural Networks (CNNs) in that specific context. In this paper, we start by investigating why plain ViT architectures exhibit a suboptimal performance when dealing with the detection of facial forgeries. Our analysis reveals that, as compared to CNNs, ViT struggles to model localized forgery artifacts that typically characterize deepfakes. Based on this observation, we propose a deepfake detection framework called FakeFormer, which extends ViTs to enforce the extraction of subtle inconsistency-prone information. For that purpose, an explicit attention learning guided by artifact-vulnerable patches and tailored to ViTs is introduced. Extensive experiments are conducted on diverse well-known datasets, including FF++, Celeb-DF, WildDeepfake, DFD, DFDCP, and DFDC. The results show that FakeFormer outperforms the state-of-the-art in terms of generalization and computational cost, without the need for large-scale training datasets. The code is available at \url{https://github.com/10Ring/FakeFormer}.