PyTorchKR Papers

“Image Classification”

총 18,403편
태그·기간 등 상세 필터

카테고리 탐색

arXiv:2306.06146

Hidden Classification Layers: Enhancing linear separability between classes in neural networks layers

In the context of classification problems, Deep Learning (DL) approaches represent state of art. Many DL approaches are based on variations of standard multi-layer feed-forward neural networks. These are also referred to as deep networks. The basic idea is that each hidden neural layer accomplishes a data transformation which is expected to make the data representation "somewhat more linearly separable" than the previous one to obtain a final data representation which is as linearly separable as possible. However, determining the appropriate neural network parameters that can perform these transformations is a critical problem. In this paper, we investigate the impact on deep network classifier performances of a training approach favouring solutions where data representations at the hidden layers have a higher degree of linear separability between the classes with respect to standard methods. To this aim, we propose a neural network architecture which induces an error function involving the outputs of all the network layers. Although similar approaches have already been partially discussed in the past literature, here we propose a new architecture with a novel error function and an extensive experimental analysis. This experimental analysis was made in the context of image classification tasks considering four widely used datasets. The results show that our approach improves the accuracy on the test set in all the considered cases.

arXiv:2404.12081공식 코드 · Papers with Code

MaskCD: A Remote Sensing Change Detection Network Based on Mask Classification

Change detection (CD) from remote sensing (RS) images using deep learning has been widely investigated in the literature. It is typically regarded as a pixel-wise labeling task that aims to classify each pixel as changed or unchanged. Although per-pixel classification networks in encoder-decoder structures have shown dominance, they still suffer from imprecise boundaries and incomplete object delineation at various scenes. For high-resolution RS images, partly or totally changed objects are more worthy of attention rather than a single pixel. Therefore, we revisit the CD task from the mask prediction and classification perspective and propose MaskCD to detect changed areas by adaptively generating categorized masks from input image pairs. Specifically, it utilizes a cross-level change representation perceiver (CLCRP) to learn multiscale change-aware representations and capture spatiotemporal relations from encoded features by exploiting deformable multihead self-attention (DeformMHSA). Subsequently, a masked-attention-based detection transformers (MA-DETR) decoder is developed to accurately locate and identify changed objects based on masked attention and self-attention mechanisms. It reconstructs the desired changed objects by decoding the pixel-wise representations into learnable mask proposals and making final predictions from these candidates. Experimental results on five benchmark datasets demonstrate the proposed approach outperforms other state-of-the-art models. Codes and pretrained models are available online (https://github.com/EricYu97/MaskCD).

arXiv:2501.09278

Text-guided Synthetic Geometric Augmentation for Zero-shot 3D Understanding

Zero-shot recognition models require extensive training data for generalization. However, in zero-shot 3D classification, collecting 3D data and captions is costly and laborintensive, posing a significant barrier compared to 2D vision. Recent advances in generative models have achieved unprecedented realism in synthetic data production, and recent research shows the potential for using generated data as training data. Here, naturally raising the question: Can synthetic 3D data generated by generative models be used as expanding limited 3D datasets? In response, we present a synthetic 3D dataset expansion method, Textguided Geometric Augmentation (TeGA). TeGA is tailored for language-image-3D pretraining, which achieves SoTA in zero-shot 3D classification, and uses a generative textto-3D model to enhance and extend limited 3D datasets. Specifically, we automatically generate text-guided synthetic 3D data and introduce a consistency filtering strategy to discard noisy samples where semantics and geometric shapes do not match with text. In the experiment to double the original dataset size using TeGA, our approach demonstrates improvements over the baselines, achieving zeroshot performance gains of 3.0% on Objaverse-LVIS, 4.6% on ScanObjectNN, and 8.7% on ModelNet40. These results demonstrate that TeGA effectively bridges the 3D data gap, enabling robust zero-shot 3D classification even with limited real training data and paving the way for zero-shot 3D vision application.

arXiv:2607.16030v1quant-ph

양자 Fisher 정보를 활용한 양자 지속 학습 재고

Rethinking Quantum Continual Learning with Quantum Fisher Information

양자 지속 학습은 순차적 과제에서 양자 모델을 학습하면서 이전에 학습한 지식을 잃지 않는 것을 목표로 한다. 그러나 변분 양자 분류기(VQC)는 비정상 과제 분포 하에서 치명적 망각(catastrophic forgetting)에 취약하다. 우리는 망각 완화를 위한 양자 Fisher 정보(QFI) 기반 정규화 방법인 양자 탄성 가중치 통합(QEWC)을 제안한다. 측정에 의존하는 출력 통계를 통해 매개변수 중요도를 측정하는 고전 Fisher 정보(CFI) 기반의 기존 탄성 가중치 통합(EWC)과 달리, QEWC는 QFI를 사용하여 매개변수화된 양자 상태의 고유한 민감도를 정량화한다. 이는 중요한 매개변수가 양자 상태 매니폴드(manifold)의 국소적 반응으로 식별되는 정보-기하학적 관점을 제공한다. 우리는 고전적 이미지 분류 및 양자 위상 분류 과제를 포함한 순차적 이진 분류 과제에서 학습된 VQC에 대해 QEWC를 평가한다. 시뮬레이션 결과, 정규화 없이 순차 학습을 하면 심각한 망각이 발생하지만, CFI 기반 EWC와 QFI 기반 QEWC 모두 이전 과제의 유지를 개선한다. 기제 분석은 두 방법이 서로 다른 정규화 기하학을 부과함을 추가로 보여준다. CFI는 측정에 민감한 방향에 선택적으로 작용하는 반면, QFI는 매개변수 공간 전반에 더 밀집된 상태-기하학적 제약을 부과한다. 탈분극(depolarizing) 노이즈 하에서 CFI 값은 저하된 측정 통계에 의해 강하게 억제되는 반면, QFI는 노이즈가 있는 매개변수화된 양자 상태의 더 안정적인 민감도 구조를 보존한다. 이러한 결과는 QEWC를 양자 상태 기하학을 통해 양자 지속 학습에서 망각을 연구하고 완화하기 위한 물리적으로 동기 부여된 접근법으로 확립한다.

arXiv:1809.10801

Effective Cloud Detection and Segmentation using a Gradient-Based Algorithm for Satellite Imagery; Application to improve PERSIANN-CCS

Being able to effectively identify clouds and monitor their evolution is one important step toward more accurate quantitative precipitation estimation and forecast. In this study, a new gradient-based cloud-image segmentation technique is developed using tools from image processing techniques. This method integrates morphological image gradient magnitudes to separable cloud systems and patches boundaries. A varying scale-kernel is implemented to reduce the sensitivity of image segmentation to noise and capture objects with various finenesses of the edges in remote-sensing images. The proposed method is flexible and extendable from single- to multi-spectral imagery. Case studies were carried out to validate the algorithm by applying the proposed segmentation algorithm to synthetic radiances for channels of the Geostationary Operational Environmental Satellites (GOES-R) simulated by a high-resolution weather prediction model. The proposed method compares favorably with the existing cloud-patch-based segmentation technique implemented in the PERSIANN-CCS (Precipitation Estimation from Remotely Sensed Information using Artificial Neural Network - Cloud Classification System) rainfall retrieval algorithm. Evaluation of event-based images indicates that the proposed algorithm has potential to improve rain detection and estimation skills with an average of more than 45% gain comparing to the segmentation technique used in PERSIANN-CCS and identifying cloud regions as objects with accuracy rates up to 98%.

arXiv:2002.00185

Deeply Activated Salient Region for Instance Search

The performance of instance search depends heavily on the ability to locate and describe a wide variety of object instances in a video/image collection. Due to the lack of proper mechanism in locating instances and deriving feature representation, instance search is generally only effective for retrieving instances of known object categories. In this paper, a simple but effective instance-level feature representation is presented. Different from other approaches, the issues in class-agnostic instance localization and distinctive feature representation are considered. The former is achieved by detecting salient instance regions from an image by a layer-wise back-propagation process. The back-propagation starts from the last convolution layer of a pre-trained CNN that is originally used for classification. The back-propagation proceeds layer-by-layer until it reaches the input layer. This allows the salient instance regions in the input image from both known and unknown categories to be activated. Each activated salient region covers the full or more usually a major range of an instance. The distinctive feature representation is produced by average-pooling on the feature map of certain layer with the detected instance region. Experiments show that such kind of feature representation demonstrates considerably better performance over most of the existing approaches. In addition, we show that the proposed feature descriptor is also suitable for content-based image search.

arXiv:2106.09444

Unsupervised classification of cell imaging data using the quantization error in a Self Organizing Map

This study exploits previously demonstrated properties such as sensitivity to the spatial extent and the intensity of local image contrast of the quantization error in the output of a Self Organizing Map (SOM QE). Here, the SOM QE is applied to double color staining based cell viability data in 96 image simulations. The results show that the SOM QE consistently and in only a few seconds detects fine regular spatial increases in relative amounts of RED or GREEN pixel staining across the test images, reflecting small, systematic increases or decreases in the percentage of theoretical cell viability below the critical threshold. Such small changes may carry clinical significance, but are almost impossible to detect by human vision. Moreover, we demonstrate a clear sensitivity of the SOM QE to differences in the relative physical luminance (Y) of the colors, which here translates into a RED GREEN color selectivity. Across differences in relative luminance, the SOM QE exhibits consistently greater sensitivity to the smallest spatial increases in RED image pixels compared with smallest increases of identical spatial extents in GREEN image pixels. Further selective color contrast studies on simulations of biological imaging data will allow generating increasingly larger benchmark datasets and, ultimately, unravel the full potential of fast, economic, and unprecedentedly precise biological data analysis using the SOM QE.

arXiv:2201.09867

Importance of Preprocessing in Histopathology Image Classification Using Deep Convolutional Neural Network

The aim of this study is to propose an alternative and hybrid solution method for diagnosing the disease from histopathology images taken from animals with paratuberculosis and intact intestine. In detail, the hybrid method is based on using both image processing and deep learning for better results. Reliable disease detection from histo-pathology images is known as an open problem in medical image processing and alternative solutions need to be developed. In this context, 520 histopathology images were collected in a joint study with Burdur Mehmet Akif Ersoy University, Faculty of Veterinary Medicine, and Department of Pathology. Manually detecting and interpreting these images requires expertise and a lot of processing time. For this reason, veterinarians, especially newly recruited physicians, have a great need for imaging and computer vision systems in the development of detection and treatment methods for this disease. The proposed solution method in this study is to use the CLAHE method and image processing together. After this preprocessing, the diagnosis is made by classifying a convolutional neural network sup-ported by the VGG-16 architecture. This method uses completely original dataset images. Two types of systems were applied for the evaluation parameters. While the F1 Score was 93% in the method classified without data preprocessing, it was 98% in the method that was preprocessed with the CLAHE method.

arXiv:1805.03699

Fast and Accurate Tumor Segmentation of Histology Images using Persistent Homology and Deep Convolutional Features

Tumor segmentation in whole-slide images of histology slides is an important step towards computer-assisted diagnosis. In this work, we propose a tumor segmentation framework based on the novel concept of persistent homology profiles (PHPs). For a given image patch, the homology profiles are derived by efficient computation of persistent homology, which is an algebraic tool from homology theory. We propose an efficient way of computing topological persistence of an image, alternative to simplicial homology. The PHPs are devised to distinguish tumor regions from their normal counterparts by modeling the atypical characteristics of tumor nuclei. We propose two variants of our method for tumor segmentation: one that targets speed without compromising accuracy and the other that targets higher accuracy. The fast version is based on the selection of exemplar image patches from a convolution neural network (CNN) and patch classification by quantifying the divergence between the PHPs of exemplars and the input image patch. Detailed comparative evaluation shows that the proposed algorithm is significantly faster than competing algorithms while achieving comparable results. The accurate version combines the PHPs and high-level CNN features and employs a multi-stage ensemble strategy for image patch labeling. Experimental results demonstrate that the combination of PHPs and CNN features outperforms competing algorithms. This study is performed on two independently collected colorectal datasets containing adenoma, adenocarcinoma, signet and healthy cases. Collectively, the accurate tumor segmentation produces the highest average patch-level F1-score, as compared with competing algorithms, on malignant and healthy cases from both the datasets. Overall the proposed framework highlights the utility of persistent homology for histopathology image analysis.

arXiv:2302.08958공식 코드 · Papers with Code

Towards Unifying Medical Vision-and-Language Pre-training via Soft Prompts

Medical vision-and-language pre-training (Med-VLP) has shown promising improvements on many downstream medical tasks owing to its applicability to extracting generic representations from medical images and texts. Practically, there exist two typical types, \textit{i.e.}, the fusion-encoder type and the dual-encoder type, depending on whether a heavy fusion module is used. The former is superior at multi-modal tasks owing to the sufficient interaction between modalities; the latter is good at uni-modal and cross-modal tasks due to the single-modality encoding ability. To take advantage of these two types, we propose an effective yet straightforward scheme named PTUnifier to unify the two types. We first unify the input format by introducing visual and textual prompts, which serve as a feature bank that stores the most representative images/texts. By doing so, a single model could serve as a \textit{foundation model} that processes various tasks adopting different input formats (\textit{i.e.}, image-only, text-only, and image-text-pair). Furthermore, we construct a prompt pool (instead of static ones) to improve diversity and scalability. Experimental results show that our approach achieves state-of-the-art results on a broad range of tasks, spanning uni-modal tasks (\textit{i.e.}, image/text classification and text summarization), cross-modal tasks (\textit{i.e.}, image-to-text generation and image-text/text-image retrieval), and multi-modal tasks (\textit{i.e.}, visual question answering), demonstrating the effectiveness of our approach. Note that the adoption of prompts is orthogonal to most existing Med-VLP approaches and could be a beneficial and complementary extension to these approaches.

arXiv:2309.01487공식 코드 · Papers with Code

GenSelfDiff-HIS: Generative Self-Supervision Using Diffusion for Histopathological Image Segmentation

Histopathological image segmentation is a laborious and time-intensive task, often requiring analysis from experienced pathologists for accurate examinations. To reduce this burden, supervised machine-learning approaches have been adopted using large-scale annotated datasets for histopathological image analysis. However, in several scenarios, the availability of large-scale annotated data is a bottleneck while training such models. Self-supervised learning (SSL) is an alternative paradigm that provides some respite by constructing models utilizing only the unannotated data which is often abundant. The basic idea of SSL is to train a network to perform one or many pseudo or pretext tasks on unannotated data and use it subsequently as the basis for a variety of downstream tasks. It is seen that the success of SSL depends critically on the considered pretext task. While there have been many efforts in designing pretext tasks for classification problems, there haven't been many attempts on SSL for histopathological segmentation. Motivated by this, we propose an SSL approach for segmenting histopathological images via generative diffusion models in this paper. Our method is based on the observation that diffusion models effectively solve an image-to-image translation task akin to a segmentation task. Hence, we propose generative diffusion as the pretext task for histopathological image segmentation. We also propose a multi-loss function-based fine-tuning for the downstream task. We validate our method using several metrics on two publically available datasets along with a newly proposed head and neck (HN) cancer dataset containing hematoxylin and eosin (H\&E) stained images along with annotations. Codes will be made public at https://github.com/suhas-srinath/GenSelfDiff-HIS.

arXiv:2310.03432

Mitigating the Influence of Domain Shift in Skin Lesion Classification: A Benchmark Study of Unsupervised Domain Adaptation Methods on Dermoscopic Images

The potential of deep neural networks in skin lesion classification has already been demonstrated to be on-par if not superior to the dermatologists diagnosis. However, the performance of these models usually deteriorates when the test data differs significantly from the training data (i.e. domain shift). This concerning limitation for models intended to be used in real-world skin lesion classification tasks poses a risk to patients. For example, different image acquisition systems or previously unseen anatomical sites on the patient can suffice to cause such domain shifts. Mitigating the negative effect of such shifts is therefore crucial, but developing effective methods to address domain shift has proven to be challenging. In this study, we carry out an in-depth analysis of eight different unsupervised domain adaptation methods to analyze their effectiveness in improving generalization for dermoscopic datasets. To ensure robustness of our findings, we test each method on a total of ten distinct datasets, thereby covering a variety of possible domain shifts. In addition, we investigated which factors in the domain shifted datasets have an impact on the effectiveness of domain adaptation methods. Our findings show that all of the eight domain adaptation methods result in improved AUPRC for the majority of analyzed datasets. Altogether, these results indicate that unsupervised domain adaptations generally lead to performance improvements for the binary melanoma-nevus classification task regardless of the nature of the domain shift. However, small or heavily imbalanced datasets lead to a reduced conformity of the results due to the influence of these factors on the methods performance.

arXiv:2009.10401

Dynamic Fusion based Federated Learning for COVID-19 Detection

Medical diagnostic image analysis (e.g., CT scan or X-Ray) using machine learning is an efficient and accurate way to detect COVID-19 infections. However, sharing diagnostic images across medical institutions is usually not allowed due to the concern of patients' privacy. This causes the issue of insufficient datasets for training the image classification model. Federated learning is an emerging privacy-preserving machine learning paradigm that produces an unbiased global model based on the received updates of local models trained by clients without exchanging clients' local data. Nevertheless, the default setting of federated learning introduces huge communication cost of transferring model updates and can hardly ensure model performance when data heterogeneity of clients heavily exists. To improve communication efficiency and model performance, in this paper, we propose a novel dynamic fusion-based federated learning approach for medical diagnostic image analysis to detect COVID-19 infections. First, we design an architecture for dynamic fusion-based federated learning systems to analyse medical diagnostic images. Further, we present a dynamic fusion method to dynamically decide the participating clients according to their local model performance and schedule the model fusion-based on participating clients' training time. In addition, we summarise a category of medical diagnostic image datasets for COVID-19 detection, which can be used by the machine learning community for image analysis. The evaluation results show that the proposed approach is feasible and performs better than the default setting of federated learning in terms of model performance, communication efficiency and fault tolerance.

arXiv:2104.07077

Self-supervised Learning of 3D Object Understanding by Data Association and Landmark Estimation for Image Sequence

In this paper, we propose a self-supervised learningmethod for multi-object pose estimation. 3D object under-standing from 2D image is a challenging task that infers ad-ditional dimension from reduced-dimensional information.In particular, the estimation of the 3D localization or orien-tation of an object requires precise reasoning, unlike othersimple clustering tasks such as object classification. There-fore, the scale of the training dataset becomes more cru-cial. However, it is challenging to obtain large amount of3D dataset since achieving 3D annotation is expensive andtime-consuming. If the scale of the training dataset can beincreased by involving the image sequence obtained fromsimple navigation, it is possible to overcome the scale lim-itation of the dataset and to have efficient adaptation tothe new environment. However, when the self annotation isconducted on single image by the network itself, trainingperformance of the network is bounded to the self perfor-mance. Therefore, we propose a strategy to exploit multipleobservations of the object in the image sequence in orderto surpass the self-performance: first, the landmarks for theglobal object map are estimated through network predic-tion and data association, and the corrected annotation fora single frame is obtained. Then, network fine-tuning is con-ducted including the dataset obtained by self-annotation,thereby exceeding the performance boundary of the networkitself. The proposed method was evaluated on the KITTIdriving scene dataset, and we demonstrate the performanceimprovement in the pose estimation of multi-object in 3D space.

arXiv:2303.03242

Evaluating the Fairness of Deep Learning Uncertainty Estimates in Medical Image Analysis

Although deep learning (DL) models have shown great success in many medical image analysis tasks, deployment of the resulting models into real clinical contexts requires: (1) that they exhibit robustness and fairness across different sub-populations, and (2) that the confidence in DL model predictions be accurately expressed in the form of uncertainties. Unfortunately, recent studies have indeed shown significant biases in DL models across demographic subgroups (e.g., race, sex, age) in the context of medical image analysis, indicating a lack of fairness in the models. Although several methods have been proposed in the ML literature to mitigate a lack of fairness in DL models, they focus entirely on the absolute performance between groups without considering their effect on uncertainty estimation. In this work, we present the first exploration of the effect of popular fairness models on overcoming biases across subgroups in medical image analysis in terms of bottom-line performance, and their effects on uncertainty quantification. We perform extensive experiments on three different clinically relevant tasks: (i) skin lesion classification, (ii) brain tumour segmentation, and (iii) Alzheimer's disease clinical score regression. Our results indicate that popular ML methods, such as data-balancing and distributionally robust optimization, succeed in mitigating fairness issues in terms of the model performances for some of the tasks. However, this can come at the cost of poor uncertainty estimates associated with the model predictions. This tradeoff must be mitigated if fairness models are to be adopted in medical image analysis.

arXiv:2505.16756

Representation Discrepancy Bridging Method for Remote Sensing Image-Text Retrieval

Remote Sensing Image-Text Retrieval (RSITR) plays a critical role in geographic information interpretation, disaster monitoring, and urban planning by establishing semantic associations between image and textual descriptions. Existing Parameter-Efficient Fine-Tuning (PEFT) methods for Vision-and-Language Pre-training (VLP) models typically adopt symmetric adapter structures for exploring cross-modal correlations. However, the strong discriminative nature of text modality may dominate the optimization process and inhibits image representation learning. The nonnegligible imbalanced cross-modal optimization remains a bottleneck to enhancing the model performance. To address this issue, this study proposes a Representation Discrepancy Bridging (RDB) method for the RSITR task. On the one hand, a Cross-Modal Asymmetric Adapter (CMAA) is designed to enable modality-specific optimization and improve feature alignment. The CMAA comprises a Visual Enhancement Adapter (VEA) and a Text Semantic Adapter (TSA). VEA mines fine-grained image features by Differential Attention (DA) mechanism, while TSA identifies key textual semantics through Hierarchical Attention (HA) mechanism. On the other hand, this study extends the traditional single-task retrieval framework to a dual-task optimization framework and develops a Dual-Task Consistency Loss (DTCL). The DTCL improves cross-modal alignment robustness through an adaptive weighted combination of cross-modal, classification, and exponential moving average consistency constraints. Experiments on RSICD and RSITMD datasets show that the proposed RDB method achieves a 6%-11% improvement in mR metrics compared to state-of-the-art PEFT methods and a 1.15%-2% improvement over the full fine-tuned GeoRSCLIP model.

arXiv:1603.04588

Classification with Repulsion Tensors: A Case Study on Face Recognition

We consider dimensionality reduction methods for face recognition in a supervised setting, using an image-as-matrix representation. A common procedure is to project image matrices into a smaller space in which the recognition is performed. These methods are often called "two-dimensional" in the literature and there exist counterparts that use an image-as-vector representation. When two face images are close to each other in the input space they may remain close after projection - but this is not desirable in the situation when these two images are from different classes, and this often affects the recognition performance. We extend a previously developed `repulsion Laplacean' technique based on adding terms to the objective function with the goal or creation a repulsion energy between such images in the projected space. This scheme, which relies on a repulsion graph, is generic and can be incorporated into various two-dimensional methods. It can be regarded as a multilinear generalization of the repulsion strategy by Kokiopoulou and Saad [Pattern Recog., 42 (2009), pp. 2392--2402]. Experimental results demonstrate that the proposed methodology offers significant recognition improvement relative to the underlying two-dimensional methods.

arXiv:2204.00656공식 코드 · Papers with Code

Consistency driven Sequential Transformers Attention Model for Partially Observable Scenes

Most hard attention models initially observe a complete scene to locate and sense informative glimpses, and predict class-label of a scene based on glimpses. However, in many applications (e.g., aerial imaging), observing an entire scene is not always feasible due to the limited time and resources available for acquisition. In this paper, we develop a Sequential Transformers Attention Model (STAM) that only partially observes a complete image and predicts informative glimpse locations solely based on past glimpses. We design our agent using DeiT-distilled and train it with a one-step actor-critic algorithm. Furthermore, to improve classification performance, we introduce a novel training objective, which enforces consistency between the class distribution predicted by a teacher model from a complete image and the class distribution predicted by our agent using glimpses. When the agent senses only 4% of the total image area, the inclusion of the proposed consistency loss in our training objective yields 3% and 8% higher accuracy on ImageNet and fMoW datasets, respectively. Moreover, our agent outperforms previous state-of-the-art by observing nearly 27% and 42% fewer pixels in glimpses on ImageNet and fMoW.

arXiv:2204.04588

Robust Cross-Modal Representation Learning with Progressive Self-Distillation

The learning objective of vision-language approach of CLIP does not effectively account for the noisy many-to-many correspondences found in web-harvested image captioning datasets, which contributes to its compute and data inefficiency. To address this challenge, we introduce a novel training framework based on cross-modal contrastive learning that uses progressive self-distillation and soft image-text alignments to more efficiently learn robust representations from noisy data. Our model distills its own knowledge to dynamically generate soft-alignment targets for a subset of images and captions in every minibatch, which are then used to update its parameters. Extensive evaluation across 14 benchmark datasets shows that our method consistently outperforms its CLIP counterpart in multiple settings, including: (a) zero-shot classification, (b) linear probe transfer, and (c) image-text retrieval, without incurring added computational cost. Analysis using an ImageNet-based robustness test-bed reveals that our method offers better effective robustness to natural distribution shifts compared to both ImageNet-trained models and CLIP itself. Lastly, pretraining with datasets spanning two orders of magnitude in size shows that our improvements over CLIP tend to scale with number of training examples.

arXiv:2312.01597공식 코드 · Papers with Code

SCLIP: Rethinking Self-Attention for Dense Vision-Language Inference

Recent advances in contrastive language-image pretraining (CLIP) have demonstrated strong capabilities in zero-shot classification by aligning visual representations with target text embeddings in an image level. However, in dense prediction tasks, CLIP often struggles to localize visual features within an image and fails to give accurate pixel-level predictions, which prevents it from functioning as a generalized visual foundation model. In this work, we aim to enhance CLIP's potential for semantic segmentation with minimal modifications to its pretrained models. By rethinking self-attention, we surprisingly find that CLIP can adapt to dense prediction tasks by simply introducing a novel Correlative Self-Attention (CSA) mechanism. Specifically, we replace the traditional self-attention block of CLIP vision encoder's last layer by our CSA module and reuse its pretrained projection matrices of query, key, and value, leading to a training-free adaptation approach for CLIP's zero-shot semantic segmentation. Extensive experiments show the advantage of CSA: we obtain a 38.2% average zero-shot mIoU across eight semantic segmentation benchmarks highlighted in this paper, significantly outperforming the existing SoTA's 33.9% and the vanilla CLIP's 14.1%.