Aiming at discovering and locating most distinctive objects from visual scenes, salient object detection (SOD) plays an essential role in various computer vision systems. Coming to the era of high resolution, SOD methods are facing new challenges. The major limitation of previous methods is that they try to identify the salient regions and estimate the accurate objects boundaries simultaneously with a single regression task at low-resolution. This practice ignores the inherent difference between the two difficult problems, resulting in poor detection quality. In this paper, we propose a novel deep learning framework for high-resolution SOD task, which disentangles the task into a low-resolution saliency classification network (LRSCN) and a high-resolution refinement network (HRRN). As a pixel-wise classification task, LRSCN is designed to capture sufficient semantics at low-resolution to identify the definite salient, background and uncertain image regions. HRRN is a regression task, which aims at accurately refining the saliency value of pixels in the uncertain region to preserve a clear object boundary at high-resolution with limited GPU memory. It is worth noting that by introducing uncertainty into the training process, our HRRN can well address the high-resolution refinement task without using any high-resolution training data. Extensive experiments on high-resolution saliency datasets as well as some widely used saliency benchmarks show that the proposed method achieves superior performance compared to the state-of-the-art methods.
Recent advancements in general-purpose or domain-specific multimodal large language models (LLMs) have witnessed remarkable progress for medical decision-making. However, they are designated for specific classification or generative tasks, and require model training or finetuning on large-scale datasets with sizeable parameters and tremendous computing, hindering their clinical utility across diverse resource-constrained scenarios in practice. In this paper, we propose a novel and lightweight framework Med-MoE (Mixture-of-Experts) that tackles both discriminative and generative multimodal medical tasks. The learning of Med-MoE consists of three steps: multimodal medical alignment, instruction tuning and routing, and domain-specific MoE tuning. After aligning multimodal medical images with LLM tokens, we then enable the model for different multimodal medical tasks with instruction tuning, together with a trainable router tailored for expert selection across input modalities. Finally, the model is tuned by integrating the router with multiple domain-specific experts, which are selectively activated and further empowered by meta expert. Comprehensive experiments on both open- and close-end medical question answering (Med-VQA) and image classification tasks across datasets such as VQA-RAD, SLAKE and Path-VQA demonstrate that our model can achieve performance superior to or on par with state-of-the-art baselines, while only requiring approximately 30\%-50\% of activated model parameters. Extensive analysis and ablations corroborate the effectiveness and practical utility of our method.
Recently computer-aided diagnosis has demonstrated promising performance, effectively alleviating the workload of clinicians. However, the inherent sample imbalance among different diseases leads algorithms biased to the majority categories, leading to poor performance for rare categories. Existing works formulated this challenge as a long-tailed problem and attempted to tackle it by decoupling the feature representation and classification. Yet, due to the imbalanced distribution and limited samples from tail classes, these works are prone to biased representation learning and insufficient classifier calibration. To tackle these problems, we propose a new Long-tailed Medical Diagnosis (LMD) framework for balanced medical image classification on long-tailed datasets. In the initial stage, we develop a Relation-aware Representation Learning (RRL) scheme to boost the representation ability by encouraging the encoder to capture intrinsic semantic features through different data augmentations. In the subsequent stage, we propose an Iterative Classifier Calibration (ICC) scheme to calibrate the classifier iteratively. This is achieved by generating a large number of balanced virtual features and fine-tuning the encoder using an Expectation-Maximization manner. The proposed ICC compensates for minority categories to facilitate unbiased classifier optimization while maintaining the diagnostic knowledge in majority classes. Comprehensive experiments on three public long-tailed medical datasets demonstrate that our LMD framework significantly surpasses state-of-the-art approaches. The source code can be accessed at https://github.com/peterlipan/LMD.
We present multiresolution tree-structured networks to process point clouds for 3D shape understanding and generation tasks. Our network represents a 3D shape as a set of locality-preserving 1D ordered list of points at multiple resolutions. This allows efficient feed-forward processing through 1D convolutions, coarse-to-fine analysis through a multi-grid architecture, and it leads to faster convergence and small memory footprint during training. The proposed tree-structured encoders can be used to classify shapes and outperform existing point-based architectures on shape classification benchmarks, while tree-structured decoders can be used for generating point clouds directly and they outperform existing approaches for image-to-shape inference tasks learned using the ShapeNet dataset. Our model also allows unsupervised learning of point-cloud based shapes by using a variational autoencoder, leading to higher-quality generated shapes.
Diagnosing different retinal diseases from Spectral Domain Optical Coherence Tomography (SD-OCT) images is a challenging task. Different automated approaches such as image processing, machine learning and deep learning algorithms have been used for early detection and diagnosis of retinal diseases. Unfortunately, these are prone to error and computational inefficiency, which requires further intervention from human experts. In this paper, we propose a novel convolution neural network architecture to successfully distinguish between different degeneration of retinal layers and their underlying causes. The proposed novel architecture outperforms other classification models while addressing the issue of gradient explosion. Our approach reaches near perfect accuracy of 99.8% and 100% for two separately available Retinal SD-OCT data-set respectively. Additionally, our architecture predicts retinal diseases in real time while outperforming human diagnosticians.
임상 응용 분야에서 신경망(neural network)은 입력 영상에서 가장 중요한 영역에 집중하고 이를 강조할 수 있어야 합니다. 소프트 어텐션(Soft-Attention) 메커니즘은 신경망이 이러한 목표를 달성할 수 있도록 합니다. 본 논문에서는 심층 신경망 구조에서 소프트 어텐션의 효과를 조사합니다. 소프트 어텐션의 핵심 목적은 중요한 특징의 가치를 높이고 노이즈를 유발하는 특징을 억제하는 것입니다. 저자들은 피부 병변을 분류(classification)할 때 소프트 어텐션 메커니즘을 적용한 경우와 적용하지 않은 경우의 VGG, ResNet, InceptionResNetv2, DenseNet 구조의 성능을 비교합니다. 기존 네트워크에 소프트 어텐션을 결합했을 때 HAM10000 데이터셋(dataset) [25]에서 93.7%의 정밀도(precision)를 달성하여 기준선(baseline)[16]보다 4.7% 향상된 성능을 보였습니다. 또한, 소프트 어텐션 결합을 통해 ISIC-2017 데이터셋 [2]에서 기준선[31] 대비 민감도 점수를 3.8% 개선하며 91.6%를 달성했습니다. 코드는 GitHub에서 공개되어 있습니다.
Knowledge distillation constitutes a simple yet effective way to improve the performance of a compact student network by exploiting the knowledge of a more powerful teacher. Nevertheless, the knowledge distillation literature remains limited to the scenario where the student and the teacher tackle the same task. Here, we investigate the problem of transferring knowledge not only across architectures but also across tasks. To this end, we study the case of object detection and, instead of following the standard detector-to-detector distillation approach, introduce a classifier-to-detector knowledge transfer framework. In particular, we propose strategies to exploit the classification teacher to improve both the detector's recognition accuracy and localization performance. Our experiments on several detectors with different backbones demonstrate the effectiveness of our approach, allowing us to outperform the state-of-the-art detector-to-detector distillation methods.
Randomized smoothing is sound when using infinite precision. However, we show that randomized smoothing is no longer sound for limited floating-point precision. We present a simple example where randomized smoothing certifies a radius of $1.26$ around a point, even though there is an adversarial example in the distance $0.8$ and extend this example further to provide false certificates for CIFAR10. We discuss the implicit assumptions of randomized smoothing and show that they do not apply to generic image classification models whose smoothed versions are commonly certified. In order to overcome this problem, we propose a sound approach to randomized smoothing when using floating-point precision with essentially equal speed and matching the certificates of the standard, unsound practice for standard classifiers tested so far. Our only assumption is that we have access to a fair coin.
This work addresses meta-learning (ML) by considering deep networks with stochastic local winner-takes-all (LWTA) activations. This type of network units results in sparse representations from each model layer, as the units are organized into blocks where only one unit generates a non-zero output. The main operating principle of the introduced units rely on stochastic principles, as the network performs posterior sampling over competing units to select the winner. Therefore, the proposed networks are explicitly designed to extract input data representations of sparse stochastic nature, as opposed to the currently standard deterministic representation paradigm. Our approach produces state-of-the-art predictive accuracy on few-shot image classification and regression experiments, as well as reduced predictive error on an active learning setting; these improvements come with an immensely reduced computational cost.
This paper introduces a novel approach to enhance the performance of the stochastic gradient descent (SGD) algorithm by incorporating a modified decay step size based on $\frac{1}{\sqrt{t}}$. The proposed step size integrates a logarithmic term, leading to the selection of smaller values in the final iterations. Our analysis establishes a convergence rate of $O(\frac{\ln T}{\sqrt{T}})$ for smooth non-convex functions without the Polyak-{\L}ojasiewicz condition. To evaluate the effectiveness of our approach, we conducted numerical experiments on image classification tasks using the FashionMNIST, and CIFAR10 datasets, and the results demonstrate significant improvements in accuracy, with enhancements of $0.5\%$ and $1.4\%$ observed, respectively, compared to the traditional $\frac{1}{\sqrt{t}}$ step size. The source code can be found at \\\url{https://github.com/Shamaeem/LNSQRTStepSize}.
Unpaired image-to-image translation using Generative Adversarial Networks (GAN) is successful in converting images among multiple domains. Moreover, recent studies have shown a way to diversify the outputs of the generator. However, since there are no restrictions on how the generator diversifies the results, it is likely to translate some unexpected features. In this paper, we propose Style-Restricted GAN (SRGAN) to demonstrate the importance of controlling the encoded features used in style diversifying process. More specifically, instead of KL divergence loss, we adopt three new losses to restrict the distribution of the encoded features: batch KL divergence loss, correlation loss, and histogram imitation loss. Further, the encoder is pre-trained with classification tasks before being used in translation process. The study reports quantitative as well as qualitative results with Precision, Recall, Density, and Coverage. The proposed three losses lead to the enhancement of the level of diversity compared to the conventional KL loss. In particular, SRGAN is found to be successful in translating with higher diversity and without changing the class-unrelated features in the CelebA face dataset. To conclude, the importance of the encoded features being well-regulated was proven with two experiments. Our implementation is available at https://github.com/shinshoji01/Style-Restricted_GAN.
This article aims to use graphic engines to simulate a large number of training data that have free annotations and possibly strongly resemble to real-world data. Between synthetic and real, a two-level domain gap exists, involving content level and appearance level. While the latter is concerned with appearance style, the former problem arises from a different mechanism, i.e, content mismatch in attributes such as camera viewpoint, object placement and lighting conditions. In contrast to the widely-studied appearance-level gap, the content-level discrepancy has not been broadly studied. To address the content-level misalignment, we propose an attribute descent approach that automatically optimizes engine attributes to enable synthetic data to approximate real-world data. We verify our method on object-centric tasks, wherein an object takes up a major portion of an image. In these tasks, the search space is relatively small, and the optimization of each attribute yields sufficiently obvious supervision signals. We collect a new synthetic asset VehicleX, and reformat and reuse existing the synthetic assets ObjectX and PersonX. Extensive experiments on image classification and object re-identification confirm that adapted synthetic data can be effectively used in three scenarios: training with synthetic data only, training data augmentation and numerically understanding dataset content.
Recent studies have shown that many deep metric learning loss functions perform very similarly under the same experimental conditions. One potential reason for this unexpected result is that all losses let the network focus on similar image regions or properties. In this paper, we investigate this by conducting a two-step analysis to extract and compare the learned visual features of the same model architecture trained with different loss functions: First, we compare the learned features on the pixel level by correlating saliency maps of the same input images. Second, we compare the clustering of embeddings for several image properties, e.g. object color or illumination. To provide independent control over these properties, photo-realistic 3D car renders similar to images in the Cars196 dataset are generated. In our analysis, we compare 14 pretrained models from a recent study and find that, even though all models perform similarly, different loss functions can guide the model to learn different features. We especially find differences between classification and ranking based losses. Our analysis also shows that some seemingly irrelevant properties can have significant influence on the resulting embedding. We encourage researchers from the deep metric learning community to use our methods to get insights into the features learned by their proposed methods.
Adversarial Training (AT), which is commonly accepted as one of the most effective approaches defending against adversarial examples, can largely harm the standard performance, thus has limited usefulness on industrial-scale production and applications. Surprisingly, this phenomenon is totally opposite in Natural Language Processing (NLP) task, where AT can even benefit for generalization. We notice the merit of AT in NLP tasks could derive from the discrete and symbolic input space. For borrowing the advantage from NLP-style AT, we propose Discrete Adversarial Training (DAT). DAT leverages VQGAN to reform the image data to discrete text-like inputs, i.e. visual words. Then it minimizes the maximal risk on such discrete images with symbolic adversarial perturbations. We further give an explanation from the perspective of distribution to demonstrate the effectiveness of DAT. As a plug-and-play technique for enhancing the visual representation, DAT achieves significant improvement on multiple tasks including image classification, object detection and self-supervised learning. Especially, the model pre-trained with Masked Auto-Encoding (MAE) and fine-tuned by our DAT without extra data can get 31.40 mCE on ImageNet-C and 32.77% top-1 accuracy on Stylized-ImageNet, building the new state-of-the-art. The code will be available at https://github.com/alibaba/easyrobust.
In the fashion domain, there exists a variety of vision-and-language (V+L) tasks, including cross-modal retrieval, text-guided image retrieval, multi-modal classification, and image captioning. They differ drastically in each individual input/output format and dataset size. It has been common to design a task-specific model and fine-tune it independently from a pre-trained V+L model (e.g., CLIP). This results in parameter inefficiency and inability to exploit inter-task relatedness. To address such issues, we propose a novel FAshion-focused Multi-task Efficient learning method for Vision-and-Language tasks (FAME-ViL) in this work. Compared with existing approaches, FAME-ViL applies a single model for multiple heterogeneous fashion tasks, therefore being much more parameter-efficient. It is enabled by two novel components: (1) a task-versatile architecture with cross-attention adapters and task-specific adapters integrated into a unified V+L model, and (2) a stable and effective multi-task training strategy that supports learning from heterogeneous data and prevents negative transfer. Extensive experiments on four fashion tasks show that our FAME-ViL can save 61.5% of parameters over alternatives, while significantly outperforming the conventional independently trained single-task models. Code is available at https://github.com/BrandonHanx/FAME-ViL.
Computer vision has become increasingly prevalent in solving real-world problems across diverse domains, including smart agriculture, fishery, and livestock management. These applications may not require processing many image frames per second, leading practitioners to use single board computers (SBCs). Although many lightweight networks have been developed for mobile/edge devices, they primarily target smartphones with more powerful processors and not SBCs with the low-end CPUs. This paper introduces a CNN-ViT hybrid network called SBCFormer, which achieves high accuracy and fast computation on such low-end CPUs. The hardware constraints of these CPUs make the Transformer's attention mechanism preferable to convolution. However, using attention on low-end CPUs presents a challenge: high-resolution internal feature maps demand excessive computational resources, but reducing their resolution results in the loss of local image details. SBCFormer introduces an architectural design to address this issue. As a result, SBCFormer achieves the highest trade-off between accuracy and speed on a Raspberry Pi 4 Model B with an ARM-Cortex A72 CPU. For the first time, it achieves an ImageNet-1K top-1 accuracy of around 80% at a speed of 1.0 frame/sec on the SBC. Code is available at https://github.com/xyongLu/SBCFormer.
With the growing size of pre-trained models, full fine-tuning and storing all the parameters for various downstream tasks is costly and infeasible. In this paper, we propose a new parameter-efficient fine-tuning method, Gradient-based Parameter Selection (GPS), demonstrating that only tuning a few selected parameters from the pre-trained model while keeping the remainder of the model frozen can generate similar or better performance compared with the full model fine-tuning method. Different from the existing popular and state-of-the-art parameter-efficient fine-tuning approaches, our method does not introduce any additional parameters and computational costs during both the training and inference stages. Another advantage is the model-agnostic and non-destructive property, which eliminates the need for any other design specific to a particular model. Compared with the full fine-tuning, GPS achieves 3.33% (91.78% vs. 88.45%, FGVC) and 9.61% (73.1% vs. 65.57%, VTAB) improvement of the accuracy with tuning only 0.36% parameters of the pre-trained model on average over 24 image classification tasks; it also demonstrates a significant improvement of 17% and 16.8% in mDice and mIoU, respectively, on medical image segmentation task. Moreover, GPS achieves state-of-the-art performance compared with existing PEFT methods.
Depth estimation is a traditional computer vision task, which plays a crucial role in understanding 3D scene geometry. Recently, deep-convolutional-neural-networks based methods have achieved promising results in the monocular depth estimation field. Specifically, the framework that combines the multi-scale features extracted by the dilated convolution based block (atrous spatial pyramid pooling, ASPP) has gained the significant improvement in the dense labeling task. However, the discretized and predefined dilation rates cannot capture the continuous context information that differs in diverse scenes and easily introduce the grid artifacts in depth estimation. In this paper, we propose an attention-based context aggregation network (ACAN) to tackle these difficulties. Based on the self-attention model, ACAN adaptively learns the task-specific similarities between pixels to model the context information. First, we recast the monocular depth estimation as a dense labeling multi-class classification problem. Then we propose a soft ordinal inference to transform the predicted probabilities to continuous depth values, which can reduce the discretization error (about 1% decrease in RMSE). Second, the proposed ACAN aggregates both the image-level and pixel-level context information for depth estimation, where the former expresses the statistical characteristic of the whole image and the latter extracts the long-range spatial dependencies for each pixel. Third, for further reducing the inconsistency between the RGB image and depth map, we construct an attention loss to minimize their information entropy. We evaluate on public monocular depth-estimation benchmark datasets (including NYU Depth V2, KITTI). The experiments demonstrate the superiority of our proposed ACAN and achieve the competitive results with the state of the arts.
Can we automatically design a Convolutional Network (ConvNet) with the highest image classification accuracy under the runtime constraint of a mobile device? Neural architecture search (NAS) has revolutionized the design of hardware-efficient ConvNets by automating this process. However, the NAS problem remains challenging due to the combinatorially large design space, causing a significant searching time (at least 200 GPU-hours). To alleviate this complexity, we propose Single-Path NAS, a novel differentiable NAS method for designing hardware-efficient ConvNets in less than 4 hours. Our contributions are as follows: 1. Single-path search space: Compared to previous differentiable NAS methods, Single-Path NAS uses one single-path over-parameterized ConvNet to encode all architectural decisions with shared convolutional kernel parameters, hence drastically decreasing the number of trainable parameters and the search cost down to few epochs. 2. Hardware-efficient ImageNet classification: Single-Path NAS achieves 74.96% top-1 accuracy on ImageNet with 79ms latency on a Pixel 1 phone, which is state-of-the-art accuracy compared to NAS methods with similar constraints (<80ms). 3. NAS efficiency: Single-Path NAS search cost is only 8 epochs (30 TPU-hours), which is up to 5,000x faster compared to prior work. 4. Reproducibility: Unlike all recent mobile-efficient NAS methods which only release pretrained models, we open-source our entire codebase at: https://github.com/dstamoulis/single-path-nas.
Recurrent neural networks are the foundation of many sequence-to-sequence models in machine learning, such as machine translation and speech synthesis. In contrast, applied quantum computing is in its infancy. Nevertheless there already exist quantum machine learning models such as variational quantum eigensolvers which have been used successfully e.g. in the context of energy minimization tasks. In this work we construct a quantum recurrent neural network (QRNN) with demonstrable performance on non-trivial tasks such as sequence learning and integer digit classification. The QRNN cell is built from parametrized quantum neurons, which, in conjunction with amplitude amplification, create a nonlinear activation of polynomials of its inputs and cell state, and allow the extraction of a probability distribution over predicted classes at each step. To study the model's performance, we provide an implementation in pytorch, which allows the relatively efficient optimization of parametrized quantum circuits with thousands of parameters. We establish a QRNN training setup by benchmarking optimization hyperparameters, and analyse suitable network topologies for simple memorisation and sequence prediction tasks from Elman's seminal paper (1990) on temporal structure learning. We then proceed to evaluate the QRNN on MNIST classification, both by feeding the QRNN each image pixel-by-pixel; and by utilising modern data augmentation as preprocessing step. Finally, we analyse to what extent the unitary nature of the network counteracts the vanishing gradient problem that plagues many existing quantum classifiers and classical RNNs.