We show how to augment any convolutional network with an attention-based global map to achieve non-local reasoning. We replace the final average pooling by an attention-based aggregation layer akin to a single transformer block, that weights how the patches are involved in the classification decision. We plug this learned aggregation layer with a simplistic patch-based convolutional network parametrized by 2 parameters (width and depth). In contrast with a pyramidal design, this architecture family maintains the input patch resolution across all the layers. It yields surprisingly competitive trade-offs between accuracy and complexity, in particular in terms of memory consumption, as shown by our experiments on various computer vision tasks: object classification, image segmentation and detection.
Random data sketching (or projection) is now a classical technique enabling, for instance, approximate numerical linear algebra and machine learning algorithms with reduced computational complexity and memory. In this context, the possibility of performing data processing (such as pattern detection or classification) directly in the sketched domain without accessing the original data was previously achieved for linear random sketching methods and compressive sensing. In this work, we show how to estimate simple signal processing tasks (such as deducing local variations in a image) directly using random quadratic projections achieved by an optical processing unit. The same approach allows for naive data classification methods directly operated in the sketched domain. We report several experiments confirming the power of our approach.
In recent years, with the development of cloud computing platforms, privacy-preserving methods for deep learning have become an urgent problem. NeuraCrypt is a private random neural network for privacy-preserving that allows data owners to encrypt the medical data before the data uploading, and data owners can train and then test their models in a cloud server with the encrypted data directly. However, we point out that the performance of NeuraCrypt is heavily degraded when using color images. In this paper, we propose a Color-NeuraCrypt to solve this problem. Experiment results show that our proposed Color-NeuraCrypt can achieve a better classification accuracy than the original one and other privacy-preserving methods.
Numerous maritime applications rely on the ability to recognize acoustic targets using passive sonar. While there is a growing reliance on pre-trained models for classification tasks, these models often require extensive computational resources and may not perform optimally when transferred to new domains due to dataset variations. To address these challenges, this work adapts the neural edge histogram descriptors (NEHD) method originally developed for image classification, to classify passive sonar signals. We conduct a comprehensive evaluation of statistical and structural texture features, demonstrating that their combination achieves competitive performance with large pre-trained models. The proposed NEHD-based approach offers a lightweight and efficient solution for underwater target recognition, significantly reducing computational costs while maintaining accuracy.
High-resolution representations (HR) are essential for dense prediction tasks such as segmentation, detection, and pose estimation. Learning HR representations is typically ignored in previous Neural Architecture Search (NAS) methods that focus on image classification. This work proposes a novel NAS method, called HR-NAS, which is able to find efficient and accurate networks for different tasks, by effectively encoding multiscale contextual information while maintaining high-resolution representations. In HR-NAS, we renovate the NAS search space as well as its searching strategy. To better encode multiscale image contexts in the search space of HR-NAS, we first carefully design a lightweight transformer, whose computational complexity can be dynamically changed with respect to different objective functions and computation budgets. To maintain high-resolution representations of the learned networks, HR-NAS adopts a multi-branch architecture that provides convolutional encoding of multiple feature resolutions, inspired by HRNet. Last, we proposed an efficient fine-grained search strategy to train HR-NAS, which effectively explores the search space, and finds optimal architectures given various tasks and computation resources. HR-NAS is capable of achieving state-of-the-art trade-offs between performance and FLOPs for three dense prediction tasks and an image classification task, given only small computational budgets. For example, HR-NAS surpasses SqueezeNAS that is specially designed for semantic segmentation while improving efficiency by 45.9%. Code is available at https://github.com/dingmyu/HR-NAS
Heterogeneous data is endemic due to the use of diverse models and settings of devices by hospitals in the field of medical imaging. However, there are few open-source frameworks for federated heterogeneous medical image analysis with personalization and privacy protection simultaneously without the demand to modify the existing model structures or to share any private data. In this paper, we proposed PPPML-HMI, an open-source learning paradigm for personalized and privacy-preserving federated heterogeneous medical image analysis. To our best knowledge, personalization and privacy protection were achieved simultaneously for the first time under the federated scenario by integrating the PerFedAvg algorithm and designing our novel cyclic secure aggregation with the homomorphic encryption algorithm. To show the utility of PPPML-HMI, we applied it to a simulated classification task namely the classification of healthy people and patients from the RAD-ChestCT Dataset, and one real-world segmentation task namely the segmentation of lung infections from COVID-19 CT scans. For the real-world task, PPPML-HMI achieved $\sim$5\% higher Dice score on average compared to conventional FL under the heterogeneous scenario. Meanwhile, we applied the improved deep leakage from gradients to simulate adversarial attacks and showed the solid privacy-preserving capability of PPPML-HMI. By applying PPPML-HMI to both tasks with different neural networks, a varied number of users, and sample sizes, we further demonstrated the strong robustness of PPPML-HMI.
The recent development of deep learning methods applied to vision has enabled their increasing integration into real-world applications to perform complex Computer Vision (CV) tasks. However, image acquisition conditions have a major impact on the performance of high-level image processing. A possible solution to overcome these limitations is to artificially augment the training databases or to design deep learning models that are robust to signal distortions. We opt here for the first solution by enriching the database with complex and realistic distortions which were ignored until now in the existing databases. To this end, we built a new versatile database derived from the well-known MS-COCO database to which we applied local and global photo-realistic distortions. These new local distortions are generated by considering the scene context of the images that guarantees a high level of photo-realism. Distortions are generated by exploiting the depth information of the objects in the scene as well as their semantics. This guarantees a high level of photo-realism and allows to explore real scenarios ignored in conventional databases dedicated to various CV applications. Our versatile database offers an efficient solution to improve the robustness of various CV tasks such as Object Detection (OD), scene segmentation, and distortion-type classification methods. The image database, scene classification index, and distortion generation codes are publicly available \footnote{\url{https://github.com/Aymanbegh/CD-COCO}}
Computer vision plays a crucial role in Advanced Assistance Systems. Most computer vision systems are based on Deep Convolutional Neural Networks (deep CNN) architectures. However, the high computational resource to run a CNN algorithm is demanding. Therefore, the methods to speed up computation have become a relevant research issue. Even though several works on architecture reduction found in the literature have not yet been achieved satisfactory results for embedded real-time system applications. This paper presents an alternative approach based on the Multilinear Feature Space (MFS) method resorting to transfer learning from large CNN architectures. The proposed method uses CNNs to generate feature maps, although it does not work as complexity reduction approach. After the training process, the generated features maps are used to create vector feature space. We use this new vector space to make projections of any new sample to classify them. Our method, named AMFC, uses the transfer learning from pre-trained CNN to reduce the classification time of new sample image, with minimal accuracy loss. Our method uses the VGG-16 model as the base CNN architecture for experiments; however, the method works with any similar CNN model. Using the well-known Vehicle Image Database and the German Traffic Sign Recognition Benchmark, we compared the classification time of the original VGG-16 model with the AMFC method, and our method is, on average, 17 times faster. The fast classification time reduces the computational and memory demands in embedded applications requiring a large CNN architecture.
Sparse decomposition has been widely used for different applications, such as source separation, image classification, image denoising and more. This paper presents a new algorithm for segmentation of an image into background and foreground text and graphics using sparse decomposition and total variation minimization. The proposed method is designed based on the assumption that the background part of the image is smoothly varying and can be represented by a linear combination of a few smoothly varying basis functions, while the foreground text and graphics can be modeled with a sparse component overlaid on the smooth background. The background and foreground are separated using a sparse decomposition framework regularized with a few suitable regularization terms which promotes the sparsity and connectivity of foreground pixels. This algorithm has been tested on a dataset of images extracted from HEVC standard test sequences for screen content coding, and is shown to have superior performance over some prior methods, including least absolute deviation fitting, k-means clustering based segmentation in DjVu and shape primitive extraction and coding (SPEC) algorithm.
Zero-shot detection (ZSD), i.e., detection on classes not seen during training, is essential for real world detection use-cases, but remains a difficult task. Recent research attempts ZSD with detection models that output embeddings instead of direct class labels. To this aim, the output of the detection model must be aligned to a learned embedding space such as CLIP. However, this alignment is hindered by detection data sets which are expensive to produce compared to image classification annotations, and the resulting lack of category diversity in the training data. We address this challenge by leveraging the CLIP embedding space in combination with image labels from ImageNet. Our results show that image labels are able to better align the detector output to the embedding space and thus have a high potential for ZSD. Compared to only training on detection data, we see a significant gain by adding image label data of 3.3 mAP for the 65/15 split on COCO on the unseen classes, i.e., we more than double the gain of related work.
Most NLP and Computer Vision tasks are limited to scarcity of labelled data. In social media emotion classification and other related tasks, hashtags have been used as indicators to label data. With the rapid increase in emoji usage of social media, emojis are used as an additional feature for major social NLP tasks. However, this is less explored in case of multimedia posts on social media where posts are composed of both image and text. At the same time, w.e have seen a surge in the interest to incorporate domain knowledge to improve machine understanding of text. In this paper, we investigate whether domain knowledge for emoji can improve the accuracy of emotion classification task. We exploit the importance of different modalities from social media post for emotion classification task using state-of-the-art deep learning architectures. Our experiments demonstrate that the three modalities (text, emoji and images) encode different information to express emotion and therefore can complement each other. Our results also demonstrate that emoji sense depends on the textual context, and emoji combined with text encodes better information than considered separately. The highest accuracy of 71.98\% is achieved with a training data of 550k posts.
Data mixing augmentation have proved to be effective in improving the generalization ability of deep neural networks. While early methods mix samples by hand-crafted policies (e.g., linear interpolation), recent methods utilize saliency information to match the mixed samples and labels via complex offline optimization. However, there arises a trade-off between precise mixing policies and optimization complexity. To address this challenge, we propose a novel automatic mixup (AutoMix) framework, where the mixup policy is parameterized and serves the ultimate classification goal directly. Specifically, AutoMix reformulates the mixup classification into two sub-tasks (i.e., mixed sample generation and mixup classification) with corresponding sub-networks and solves them in a bi-level optimization framework. For the generation, a learnable lightweight mixup generator, Mix Block, is designed to generate mixed samples by modeling patch-wise relationships under the direct supervision of the corresponding mixed labels. To prevent the degradation and instability of bi-level optimization, we further introduce a momentum pipeline to train AutoMix in an end-to-end manner. Extensive experiments on nine image benchmarks prove the superiority of AutoMix compared with state-of-the-art in various classification scenarios and downstream tasks.
This paper introduces a novel mathematical property applicable to diverse images, referred to as FINOLA (First-Order Norm+Linear Autoregressive). FINOLA represents each image in the latent space as a first-order autoregressive process, in which each regression step simply applies a shared linear model on the normalized value of its immediate neighbor. This intriguing property reveals a mathematical invariance that transcends individual images. Expanding from image grids to continuous coordinates, we unveil the presence of two underlying partial differential equations. We validate the FINOLA property from two distinct angles: image reconstruction and self-supervised learning. Firstly, we demonstrate the ability of FINOLA to auto-regress up to a 256x256 feature map (the same resolution to the image) from a single vector placed at the center, successfully reconstructing the original image by only using three 3x3 convolution layers as decoder. Secondly, we leverage FINOLA for self-supervised learning by employing a simple masked prediction approach. Encoding a single unmasked quadrant block, we autoregressively predict the surrounding masked region. Remarkably, this pre-trained representation proves highly effective in image classification and object detection tasks, even when integrated into lightweight networks, all without the need for extensive fine-tuning. The code will be made publicly available.
We introduce SPARse Fine-grained Contrastive Alignment (SPARC), a simple method for pretraining more fine-grained multimodal representations from image-text pairs. Given that multiple image patches often correspond to single words, we propose to learn a grouping of image patches for every token in the caption. To achieve this, we use a sparse similarity metric between image patches and language tokens and compute for each token a language-grouped vision embedding as the weighted average of patches. The token and language-grouped vision embeddings are then contrasted through a fine-grained sequence-wise loss that only depends on individual samples and does not require other batch samples as negatives. This enables more detailed information to be learned in a computationally inexpensive manner. SPARC combines this fine-grained loss with a contrastive loss between global image and text embeddings to learn representations that simultaneously encode global and local information. We thoroughly evaluate our proposed method and show improved performance over competing approaches both on image-level tasks relying on coarse-grained information, e.g. classification, as well as region-level tasks relying on fine-grained information, e.g. retrieval, object detection, and segmentation. Moreover, SPARC improves model faithfulness and captioning in foundational vision-language models.
We propose a method for inferring human attributes (such as gender, hair style, clothes style, expression, action) from images of people under large variation of viewpoint, pose, appearance, articulation and occlusion. Convolutional Neural Nets (CNN) have been shown to perform very well on large scale object recognition problems. In the context of attribute classification, however, the signal is often subtle and it may cover only a small part of the image, while the image is dominated by the effects of pose and viewpoint. Discounting for pose variation would require training on very large labeled datasets which are not presently available. Part-based models, such as poselets and DPM have been shown to perform well for this problem but they are limited by shallow low-level features. We propose a new method which combines part-based models and deep learning by training pose-normalized CNNs. We show substantial improvement vs. state-of-the-art methods on challenging attribute classification tasks in unconstrained settings. Experiments confirm that our method outperforms both the best part-based methods on this problem and conventional CNNs trained on the full bounding box of the person.
Research in human-centered AI has shown the benefits of systems that can explain their predictions. Methods that allow an AI to take advice from humans in response to explanations are similarly useful. While both capabilities are well-developed for transparent learning models (e.g., linear models and GA$^2$Ms), and recent techniques (e.g., LIME and SHAP) can generate explanations for opaque models, little attention has been given to advice methods for opaque models. This paper introduces LIMEADE, the first general framework that translates both positive and negative advice (expressed using high-level vocabulary such as that employed by post-hoc explanations) into an update to an arbitrary, underlying opaque model. We demonstrate the generality of our approach with case studies on seventy real-world models across two broad domains: image classification and text recommendation. We show our method improves accuracy compared to a rigorous baseline on the image classification domains. For the text modality, we apply our framework to a neural recommender system for scientific papers on a public website; our user study shows that our framework leads to significantly higher perceived user control, trust, and satisfaction.
Data augmentation methods have been shown to be a fundamental technique to improve generalization in tasks such as image, text and audio classification. Recently, automated augmentation methods have led to further improvements on image classification and object detection leading to state-of-the-art performances. Nevertheless, little work has been done on time-series data, an area that could greatly benefit from automated data augmentation given the usually limited size of the datasets. We present two sample-adaptive automatic weighting schemes for data augmentation: the first learns to weight the contribution of the augmented samples to the loss, and the second method selects a subset of transformations based on the ranking of the predicted training loss. We validate our proposed methods on a large, noisy financial dataset and on time-series datasets from the UCR archive. On the financial dataset, we show that the methods in combination with a trading strategy lead to improvements in annualized returns of over 50$\%$, and on the time-series data we outperform state-of-the-art models on over half of the datasets, and achieve similar performance in accuracy on the others.
In this work, we present RadioTransformer, a novel visual attention-driven transformer framework, that leverages radiologists' gaze patterns and models their visuo-cognitive behavior for disease diagnosis on chest radiographs. Domain experts, such as radiologists, rely on visual information for medical image interpretation. On the other hand, deep neural networks have demonstrated significant promise in similar tasks even where visual interpretation is challenging. Eye-gaze tracking has been used to capture the viewing behavior of domain experts, lending insights into the complexity of visual search. However, deep learning frameworks, even those that rely on attention mechanisms, do not leverage this rich domain information. RadioTransformer fills this critical gap by learning from radiologists' visual search patterns, encoded as 'human visual attention regions' in a cascaded global-focal transformer framework. The overall 'global' image characteristics and the more detailed 'local' features are captured by the proposed global and focal modules, respectively. We experimentally validate the efficacy of our student-teacher approach for 8 datasets involving different disease classification tasks where eye-gaze data is not available during the inference phase. Code: https://github.com/bmi-imaginelab/radiotransformer.
In this work, we present a method for synthetic CT (sCT) generation from zero-echo-time (ZTE) MRI aimed at structural and quantitative accuracies of the image, with a particular focus on the accurate bone density value prediction. We propose a loss function that favors a spatially sparse region in the image. We harness the ability of a multi-task network to produce correlated outputs as a framework to enable localisation of region of interest (RoI) via classification, emphasize regression of values within RoI and still retain the overall accuracy via global regression. The network is optimized by a composite loss function that combines a dedicated loss from each task. We demonstrate how the multi-task network with RoI focused loss offers an advantage over other configurations of the network to achieve higher accuracy of performance. This is relevant to sCT where failure to accurately estimate high Hounsfield Unit values of bone could lead to impaired accuracy in clinical applications. We compare the dose calculation maps from the proposed sCT and the real CT in a radiation therapy treatment planning setup.
Automated region of interest detection in histopathological image analysis is a challenging and important topic with tremendous potential impact on clinical practice. The deep-learning methods used in computational pathology help us to reduce costs and increase the speed and accuracy of regions of interest detection and cancer diagnosis. In this work, we propose a patch-based region of interest detection method for melanocytic skin tumor whole-slide images. We work with a dataset that contains 165 primary melanomas and nevi Hematoxylin and Eosin whole-slide images and build a deep-learning method. The proposed method performs well on a hold-out test data set including five TCGA-SKCM slides (accuracy of 93.94\% in slide classification task and intersection over union rate of 41.27\% in the region of interest detection task), showing the outstanding performance of our model on melanocytic skin tumor. Even though we test the experiments on the skin tumor dataset, our work could also be extended to other medical image detection problems, such as various tumors' classification and prediction, to help and benefit the clinical evaluation and diagnosis of different tumors.