When a small number of poisoned samples are injected into the training dataset of a deep neural network, the network can be induced to exhibit malicious behavior during inferences, which poses potential threats to real-world applications. While they have been intensively studied in classification, backdoor attacks on semantic segmentation have been largely overlooked. Unlike classification, semantic segmentation aims to classify every pixel within a given image. In this work, we explore backdoor attacks on segmentation models to misclassify all pixels of a victim class by injecting a specific trigger on non-victim pixels during inferences, which is dubbed Influencer Backdoor Attack (IBA). IBA is expected to maintain the classification accuracy of non-victim pixels and mislead classifications of all victim pixels in every single inference and could be easily applied to real-world scenes. Based on the context aggregation ability of segmentation models, we proposed a simple, yet effective, Nearest-Neighbor trigger injection strategy. We also introduce an innovative Pixel Random Labeling strategy which maintains optimal performance even when the trigger is placed far from the victim pixels. Our extensive experiments reveal that current segmentation models do suffer from backdoor attacks, demonstrate IBA real-world applicability, and show that our proposed techniques can further increase attack performance.
We propose a technique for producing "visual explanations" for decisions from a large class of CNN-based models, making them more transparent. Our approach - Gradient-weighted Class Activation Mapping (Grad-CAM), uses the gradients of any target concept, flowing into the final convolutional layer to produce a coarse localization map highlighting important regions in the image for predicting the concept. Grad-CAM is applicable to a wide variety of CNN model-families: (1) CNNs with fully-connected layers, (2) CNNs used for structured outputs, (3) CNNs used in tasks with multimodal inputs or reinforcement learning, without any architectural changes or re-training. We combine Grad-CAM with fine-grained visualizations to create a high-resolution class-discriminative visualization and apply it to off-the-shelf image classification, captioning, and visual question answering (VQA) models, including ResNet-based architectures. In the context of image classification models, our visualizations (a) lend insights into their failure modes, (b) are robust to adversarial images, (c) outperform previous methods on localization, (d) are more faithful to the underlying model and (e) help achieve generalization by identifying dataset bias. For captioning and VQA, we show that even non-attention based models can localize inputs. We devise a way to identify important neurons through Grad-CAM and combine it with neuron names to provide textual explanations for model decisions. Finally, we design and conduct human studies to measure if Grad-CAM helps users establish appropriate trust in predictions from models and show that Grad-CAM helps untrained users successfully discern a 'stronger' nodel from a 'weaker' one even when both make identical predictions. Our code is available at https://github.com/ramprs/grad-cam/, along with a demo at http://gradcam.cloudcv.org, and a video at youtu.be/COjUB9Izk6E.
Explainable artificial intelligence is the attempt to elucidate the workings of systems too complex to be directly accessible to human cognition through suitable side-information referred to as "explanations". We present a trainable explanation module for convolutional image classifiers we call bounded logit attention (BLA). The BLA module learns to select a subset of the convolutional feature map for each input instance, which then serves as an explanation for the classifier's prediction. BLA overcomes several limitations of the instancewise feature selection method "learning to explain" (L2X) introduced by Chen et al. (2018): 1) BLA scales to real-world sized image classification problems, and 2) BLA offers a canonical way to learn explanations of variable size. Due to its modularity BLA lends itself to transfer learning setups and can also be employed as a post-hoc add-on to trained classifiers. Beyond explainability, BLA may serve as a general purpose method for differentiable approximation of subset selection. In a user study we find that BLA explanations are preferred over explanations generated by the popular (Grad-)CAM method.
Vision Transformers (ViT) have made many breakthroughs in computer vision tasks. However, considerable redundancy arises in the spatial dimension of an input image, leading to massive computational costs. Therefore, We propose a coarse-to-fine vision transformer (CF-ViT) to relieve computational burden while retaining performance in this paper. Our proposed CF-ViT is motivated by two important observations in modern ViT models: (1) The coarse-grained patch splitting can locate informative regions of an input image. (2) Most images can be well recognized by a ViT model in a small-length token sequence. Therefore, our CF-ViT implements network inference in a two-stage manner. At coarse inference stage, an input image is split into a small-length patch sequence for a computationally economical classification. If not well recognized, the informative patches are identified and further re-split in a fine-grained granularity. Extensive experiments demonstrate the efficacy of our CF-ViT. For example, without any compromise on performance, CF-ViT reduces 53% FLOPs of LV-ViT, and also achieves 2.01x throughput.
Medical visual question answering (VQA) is a challenging multimodal task, where Vision-Language Pre-training (VLP) models can effectively improve the generalization performance. However, most methods in the medical field treat VQA as an answer classification task which is difficult to transfer to practical application scenarios. Additionally, due to the privacy of medical images and the expensive annotation process, large-scale medical image-text pairs datasets for pretraining are severely lacking. In this paper, we propose a large-scale MultI-task Self-Supervised learning based framework (MISS) for medical VQA tasks. Unlike existing methods, we treat medical VQA as a generative task. We unify the text encoder and multimodal encoder and align image-text features through multi-task learning. Furthermore, we propose a Transfer-and-Caption method that extends the feature space of single-modal image datasets using Large Language Models (LLMs), enabling those traditional medical vision field task data to be applied to VLP. Experiments show that our method achieves excellent results with fewer multimodal datasets and demonstrates the advantages of generative VQA models.
Gradient inversion attacks aim to reconstruct local training data from intermediate gradients exposed in the federated learning framework. Despite successful attacks, all previous methods, starting from reconstructing a single data point and then relaxing the single-image limit to batch level, are only tested under hard label constraints. Even for single-image reconstruction, we still lack an analysis-based algorithm to recover augmented soft labels. In this work, we change the focus from enlarging batchsize to investigating the hard label constraints, considering a more realistic circumstance where label smoothing and mixup techniques are used in the training process. In particular, we are the first to initiate a novel algorithm to simultaneously recover the ground-truth augmented label and the input feature of the last fully-connected layer from single-input gradients, and provide a necessary condition for any analytical-based label recovery methods. Extensive experiments testify to the label recovery accuracy, as well as the benefits to the following image reconstruction. We believe soft labels in classification tasks are worth further attention in gradient inversion attacks.
The rise of graph-structured data such as social networks, regulatory networks, citation graphs, and functional brain networks, in combination with resounding success of deep learning in various applications, has brought the interest in generalizing deep learning models to non-Euclidean domains. In this paper, we introduce a new spectral domain convolutional architecture for deep learning on graphs. The core ingredient of our model is a new class of parametric rational complex functions (Cayley polynomials) allowing to efficiently compute spectral filters on graphs that specialize on frequency bands of interest. Our model generates rich spectral filters that are localized in space, scales linearly with the size of the input data for sparsely-connected graphs, and can handle different constructions of Laplacian operators. Extensive experimental results show the superior performance of our approach, in comparison to other spectral domain convolutional architectures, on spectral image classification, community detection, vertex classification and matrix completion tasks.
The `Internet of Things' has brought increased demand for AI-based edge computing in applications ranging from healthcare monitoring systems to autonomous vehicles. Quantization is a powerful tool to address the growing computational cost of such applications, and yields significant compression over full-precision networks. However, quantization can result in substantial loss of performance for complex image classification tasks. To address this, we propose a Principal Component Analysis (PCA) driven methodology to identify the important layers of a binary network, and design mixed-precision networks. The proposed Hybrid-Net achieves a more than 10% improvement in classification accuracy over binary networks such as XNOR-Net for ResNet and VGG architectures on CIFAR-100 and ImageNet datasets while still achieving up to 94% of the energy-efficiency of XNOR-Nets. This work furthers the feasibility of using highly compressed neural networks for energy-efficient neural computing in edge devices.
We introduce the initial release of our software Robustar, which aims to improve the robustness of vision classification machine learning models through a data-driven perspective. Building upon the recent understanding that the lack of machine learning model's robustness is the tendency of the model's learning of spurious features, we aim to solve this problem from its root at the data perspective by removing the spurious features from the data before training. In particular, we introduce a software that helps the users to better prepare the data for training image classification models by allowing the users to annotate the spurious features at the pixel level of images. To facilitate this process, our software also leverages recent advances to help identify potential images and pixels worthy of attention and to continue the training with newly annotated data. Our software is hosted at the GitHub Repository https://github.com/HaohanWang/Robustar.
We consider neural network training, in applications in which there are many possible classes, but at test-time, the task is a binary classification task of determining whether the given example belongs to a specific class, where the class of interest can be different each time the classifier is applied. For instance, this is the case for real-time image search. We define the Single Logit Classification (SLC) task: training the network so that at test-time, it would be possible to accurately identify whether the example belongs to a given class in a computationally efficient manner, based only on the output logit for this class. We propose a natural principle, the Principle of Logit Separation, as a guideline for choosing and designing losses suitable for the SLC. We show that the cross-entropy loss function is not aligned with the Principle of Logit Separation. In contrast, there are known loss functions, as well as novel batch loss functions that we propose, which are aligned with this principle. In total, we study seven loss functions. Our experiments show that indeed in almost all cases, losses that are aligned with the Principle of Logit Separation obtain at least 20% relative accuracy improvement in the SLC task compared to losses that are not aligned with it, and sometimes considerably more. Furthermore, we show that fast SLC does not cause any drop in binary classification accuracy, compared to standard classification in which all logits are computed, and yields a speedup which grows with the number of classes. For instance, we demonstrate a 10x speedup when the number of classes is 400,000. Tensorflow code for optimizing the new batch losses is publicly available at https://github.com/cruvadom/Logit Separation.
Molecular and morphological characters, as important parts of biological taxonomy, are contradictory but need to be integrated. Organism's image recognition and bioinformatics are emerging and hot problems nowadays but with a gap between them. In this work, a multi-branching recognition framework mediated by genetic information bridges this barrier, which establishes the link between macro-morphology and micro-molecular information of mushrooms. The novel multi-perspective structure is proposed to fuse the feature images from three branching models, which significantly improves the accuracy of recognition by about 10% and up to more than 90%. Further, genetic information is implemented to the mushroom image recognition task by using genetic distance embeddings as the representation space for predicting image distance and species identification. Semantic overfitting of traditional classification tasks and the granularity of fine-grained image recognition are also discussed in depth for the first time. The generalizability of the model was investigated in fine-grained scenarios using zero-shot learning tasks, which could predict the taxonomic and evolutionary information of unseen samples. We presented the first method to map images to DNA, namely used an encoder mapping image to genetic distances, and then decoded DNA through a pre-trained decoder, where the total test accuracy on 37 species for DNA prediction is 87.45%. This study creates a novel recognition framework by systematically studying the mushroom image recognition problem, bridging the gap between macroscopic biological information and microscopic molecular information, which will provide a new reference for intelligent biometrics in the future.
Groundbreaking language-vision architectures like CLIP and DALL-E proved the utility of training on large amounts of noisy image-text data, without relying on expensive accurate labels used in standard vision unimodal supervised learning. The resulting models showed capabilities of strong text-guided image generation and transfer to downstream tasks, while performing remarkably at zero-shot classification with noteworthy out-of-distribution robustness. Since then, large-scale language-vision models like ALIGN, BASIC, GLIDE, Flamingo and Imagen made further improvements. Studying the training and capabilities of such models requires datasets containing billions of image-text pairs. Until now, no datasets of this size have been made openly available for the broader research community. To address this problem and democratize research on large-scale multi-modal models, we present LAION-5B - a dataset consisting of 5.85 billion CLIP-filtered image-text pairs, of which 2.32B contain English language. We show successful replication and fine-tuning of foundational models like CLIP, GLIDE and Stable Diffusion using the dataset, and discuss further experiments enabled with an openly available dataset of this scale. Additionally we provide several nearest neighbor indices, an improved web-interface for dataset exploration and subset generation, and detection scores for watermark, NSFW, and toxic content detection. Announcement page https://laion.ai/laion-5b-a-new-era-of-open-large-scale-multi-modal-datasets/
When encountering a dubious diagnostic case, medical instance retrieval can help radiologists make evidence-based diagnoses by finding images containing instances similar to a query case from a large image database. The similarity between the query case and retrieved similar cases is determined by visual features extracted from pathologically abnormal regions. However, the manifestation of these regions often lacks specificity, i.e., different diseases can have the same manifestation, and different manifestations may occur at different stages of the same disease. To combat the manifestation ambiguity in medical instance retrieval, we propose a novel deep framework called Y-Net, encoding images into compact hash-codes generated from convolutional features by feature aggregation. Y-Net can learn highly discriminative convolutional features by unifying the pixel-wise segmentation loss and classification loss. The segmentation loss allows exploring subtle spatial differences for good spatial-discriminability while the classification loss utilizes class-aware semantic information for good semantic-separability. As a result, Y-Net can enhance the visual features in pathologically abnormal regions and suppress the disturbing of the background during model training, which could effectively embed discriminative features into the hash-codes in the retrieval stage. Extensive experiments on two medical image datasets demonstrate that Y-Net can alleviate the ambiguity of pathologically abnormal regions and its retrieval performance outperforms the state-of-the-art method by an average of 9.27\% on the returned list of 10.
Neural ordinary differential equations (neural ODEs) have emerged as a novel network architecture that bridges dynamical systems and deep learning. However, the gradient obtained with the continuous adjoint method in the vanilla neural ODE is not reverse-accurate. Other approaches suffer either from an excessive memory requirement due to deep computational graphs or from limited choices for the time integration scheme, hampering their application to large-scale complex dynamical systems. To achieve accurate gradients without compromising memory efficiency and flexibility, we present a new neural ODE framework, PNODE, based on high-level discrete adjoint algorithmic differentiation. By leveraging discrete adjoint time integrators and advanced checkpointing strategies tailored for these integrators, PNODE can provide a balance between memory and computational costs, while computing the gradients consistently and accurately. We provide an open-source implementation based on PyTorch and PETSc, one of the most commonly used portable, scalable scientific computing libraries. We demonstrate the performance through extensive numerical experiments on image classification and continuous normalizing flow problems. We show that PNODE achieves the highest memory efficiency when compared with other reverse-accurate methods. On the image classification problems, PNODE is up to two times faster than the vanilla neural ODE and up to 2.3 times faster than the best existing reverse-accurate method. We also show that PNODE enables the use of the implicit time integration methods that are needed for stiff dynamical systems.
A large amount of research effort has been dedicated to adapting boosting for imbalanced classification. However, boosting methods are yet to be satisfactorily immune to class imbalance, especially for multi-class problems. This is because most of the existing solutions for handling class imbalance rely on expensive cost set tuning for determining the proper level of compensation. We show that the assignment of weights to the component classifiers of a boosted ensemble can be thought of as a game of Tug of War between the classes in the margin space. We then demonstrate how this insight can be used to attain a good compromise between the rare and abundant classes without having to resort to cost set tuning, which has long been the norm for imbalanced classification. The solution is based on a lexicographic linear programming framework which requires two stages. Initially, class-specific component weight combinations are found so as to minimize a hinge loss individually for each of the classes. Subsequently, the final component weights are assigned so that the maximum deviation from the class-specific minimum loss values (obtained in the previous stage) is minimized. Hence, the proposal is not only restricted to two-class situations, but is also readily applicable to multi-class problems. Additionally,we also derive the dual formulation corresponding to the proposed framework. Experiments conducted on artificial and real-world imbalanced datasets as well as on challenging applications such as hyperspectral image classification and ImageNet classification establish the efficacy of the proposal.
Open-Set Classification (OSC) intends to adapt closed-set classification models to real-world scenarios, where the classifier must correctly label samples of known classes while rejecting previously unseen unknown samples. Only recently, research started to investigate on algorithms that are able to handle these unknown samples correctly. Some of these approaches address OSC by including into the training set negative samples that a classifier learns to reject, expecting that these data increase the robustness of the classifier on unknown classes. Most of these approaches are evaluated on small-scale and low-resolution image datasets like MNIST, SVHN or CIFAR, which makes it difficult to assess their applicability to the real world, and to compare them among each other. We propose three open-set protocols that provide rich datasets of natural images with different levels of similarity between known and unknown classes. The protocols consist of subsets of ImageNet classes selected to provide training and testing data closer to real-world scenarios. Additionally, we propose a new validation metric that can be employed to assess whether the training of deep learning models addresses both the classification of known samples and the rejection of unknown samples. We use the protocols to compare the performance of two baseline open-set algorithms to the standard SoftMax baseline and find that the algorithms work well on negative samples that have been seen during training, and partially on out-of-distribution detection tasks, but drop performance in the presence of samples from previously unseen unknown classes.
Developing robust deep learning models for fetal ultrasound image analysis requires comprehensive, high-quality datasets to effectively learn informative data representations within the domain. However, the scarcity of labelled ultrasound images poses substantial challenges, especially in low-resource settings. To tackle this challenge, we leverage synthetic data to enhance the generalizability of deep learning models. This study proposes a diffusion-based method, Fetal Ultrasound LoRA (FU-LoRA), which involves fine-tuning latent diffusion models using the LoRA technique to generate synthetic fetal ultrasound images. These synthetic images are integrated into a hybrid dataset that combines real-world and synthetic images to improve the performance of zero-shot classifiers in low-resource settings. Our experimental results on fetal ultrasound images from African cohorts demonstrate that FU-LoRA outperforms the baseline method by a 13.73% increase in zero-shot classification accuracy. Furthermore, FU-LoRA achieves the highest accuracy of 82.40%, the highest F-score of 86.54%, and the highest AUC of 89.78%. It demonstrates that the FU-LoRA method is effective in the zero-shot classification of fetal ultrasound images in low-resource settings. Our code and data are publicly accessible on https://github.com/13204942/FU-LoRA.
eXplainable Artificial Intelligence (XAI) has emerged as an essential requirement when dealing with mission-critical applications, ensuring transparency and interpretability of the employed black box AI models. The significance of XAI spans various domains, from healthcare to finance, where understanding the decision-making process of deep learning algorithms is essential. Most AI-based computer vision models are often black boxes; hence, providing explainability of deep neural networks in image processing is crucial for their wide adoption and deployment in medical image analysis, autonomous driving, and remote sensing applications. Recently, several XAI methods for image classification tasks have been introduced. On the contrary, image segmentation has received comparatively less attention in the context of explainability, although it is a fundamental task in computer vision applications, especially in remote sensing. Only some research proposes gradient-based XAI algorithms for image segmentation. This paper adapts the recent gradient-free Sobol XAI method for semantic segmentation. To measure the performance of the Sobol method for segmentation, we propose a quantitative XAI evaluation method based on a learnable noise model. The main objective of this model is to induce noise on the explanation maps, where higher induced noise signifies low accuracy and vice versa. A benchmark analysis is conducted to evaluate and compare performance of three XAI methods, including Seg-Grad-CAM, Seg-Grad-CAM++ and Seg-Sobol using the proposed noise-based evaluation technique. This constitutes the first attempt to run and evaluate XAI methods using high-resolution satellite images.
Convolutional Neural Networks (CNNs) for visual tasks are believed to learn both the low-level textures and high-level object attributes, throughout the network depth. This paper further investigates the `texture bias' in CNNs. To this end, we regenerate multiple instances of training examples from each original image, through a process we call `repainting'. The repainted examples preserve the shape and structure of the regions and objects within the scenes, but diversify their texture and color. Our method can regenerate a same image at different daylight, season, or weather conditions, can have colorization or de-colorization effects, or even bring back some texture information from blacked-out areas. The in-place repaint allows us to further use these repainted examples for improving the generalization of CNNs. Through an extensive set of experiments, we demonstrate the usefulness of the repainted examples in training, for the tasks of image classification (ImageNet) and object detection (COCO), over several state-of-the-art network architectures at different capacities, and across different data availability regimes.
Scaling up contrastive language-image pretraining (CLIP) is critical for empowering both vision and multimodal models. We present EVA-CLIP-18B, the largest and most powerful open-source CLIP model to date, with 18-billion parameters. With only 6-billion training samples seen, EVA-CLIP-18B achieves an exceptional 80.7% zero-shot top-1 accuracy averaged across 27 widely recognized image classification benchmarks, outperforming its forerunner EVA-CLIP (5-billion parameters) and other open-source CLIP models by a large margin. Remarkably, we observe a consistent performance improvement with the model size scaling of EVA-CLIP, despite maintaining a constant training dataset of 2-billion image-text pairs from LAION-2B and COYO-700M. This dataset is openly available and much smaller than the in-house datasets (e.g., DFN-5B, WebLI-10B) employed in other state-of-the-art CLIP models. EVA-CLIP-18B demonstrates the potential of EVA-style weak-to-strong visual model scaling. With our model weights made publicly available, we hope to facilitate future research in vision and multimodal foundation models.