Magnetic Resonance Imaging represents an important diagnostic modality; however, its inherently slow acquisition process poses challenges in obtaining fully sampled k-space data under motion in clinical scenarios such as abdominal, cardiac, and prostate imaging. In the absence of fully sampled acquisitions, which can serve as ground truth data, training deep learning algorithms in a supervised manner to predict the underlying ground truth image becomes an impossible task. To address this limitation, self-supervised methods have emerged as a viable alternative, leveraging available subsampled k-space data to train deep learning networks for MRI reconstruction. Nevertheless, these self-supervised approaches often fall short when compared to supervised methodologies. In this paper, we introduce JSSL (Joint Supervised and Self-supervised Learning), a novel training approach for deep learning-based MRI reconstruction algorithms aimed at enhancing reconstruction quality in scenarios where target dataset(s) containing fully sampled k-space measurements are unavailable. Our proposed method operates by simultaneously training a model in a self-supervised learning setting, using subsampled data from the target dataset(s), and in a supervised learning manner, utilizing data from other datasets, referred to as proxy datasets, where fully sampled k-space data is accessible. To demonstrate the efficacy of JSSL, we utilized subsampled prostate parallel MRI measurements as the target dataset, while employing fully sampled brain and knee k-space acquisitions as proxy datasets. Our results showcase a substantial improvement over conventional self-supervised training methods, thereby underscoring the effectiveness of our joint approach. We provide a theoretical motivation for JSSL and establish a practical "rule-of-thumb" for selecting the most appropriate training approach for deep MRI reconstruction.
相關內容
Modern speech processing systems rely on self-attention. Unfortunately, token mixing with self-attention takes quadratic time in the length of the speech utterance, slowing down inference as well as training and increasing memory consumption. Cheaper alternatives to self-attention for ASR have been developed, but they fail to consistently reach the same level of accuracy. This paper, therefore, proposes a novel linear-time alternative to self-attention. It summarises an utterance with the mean over vectors for all time steps. This single summary is then combined with time-specific information. We call this method "SummaryMixing". Introducing SummaryMixing in state-of-the-art ASR models makes it feasible to preserve or exceed previous speech recognition performance while lowering the training and inference times by up to 28$\%$ and reducing the memory budget by a factor of two. The benefits of SummaryMixing can also be generalized to other speech-processing tasks, such as speech understanding.
Vision Transformers (ViTs) have emerged as powerful models in the field of computer vision, delivering superior performance across various vision tasks. However, the high computational complexity poses a significant barrier to their practical applications in real-world scenarios. Motivated by the fact that not all tokens contribute equally to the final predictions and fewer tokens bring less computational cost, reducing redundant tokens has become a prevailing paradigm for accelerating vision transformers. However, we argue that it is not optimal to either only reduce inattentive redundancy by token pruning, or only reduce duplicative redundancy by token merging. To this end, in this paper we propose a novel acceleration framework, namely token Pruning & Pooling Transformers (PPT), to adaptively tackle these two types of redundancy in different layers. By heuristically integrating both token pruning and token pooling techniques in ViTs without additional trainable parameters, PPT effectively reduces the model complexity while maintaining its predictive accuracy. For example, PPT reduces over 37% FLOPs and improves the throughput by over 45% for DeiT-S without any accuracy drop on the ImageNet dataset. The code is available at //github.com/xjwu1024/PPT and //github.com/mindspore-lab/models/
Vision Transformers (ViTs) achieve superior performance on various tasks compared to convolutional neural networks (CNNs), but ViTs are also vulnerable to adversarial attacks. Adversarial training is one of the most successful methods to build robust CNN models. Thus, recent works explored new methodologies for adversarial training of ViTs based on the differences between ViTs and CNNs, such as better training strategies, preventing attention from focusing on a single block, or discarding low-attention embeddings. However, these methods still follow the design of traditional supervised adversarial training, limiting the potential of adversarial training on ViTs. This paper proposes a novel defense method, MIMIR, which aims to build a different adversarial training methodology by utilizing Masked Image Modeling at pre-training. We create an autoencoder that accepts adversarial examples as input but takes the clean examples as the modeling target. Then, we create a mutual information (MI) penalty following the idea of the Information Bottleneck. Among the two information source inputs and corresponding adversarial perturbation, the perturbation information is eliminated due to the constraint of the modeling target. Next, we provide a theoretical analysis of MIMIR using the bounds of the MI penalty. We also design two adaptive attacks when the adversary is aware of the MIMIR defense and show that MIMIR still performs well. The experimental results show that MIMIR improves (natural and adversarial) accuracy on average by 4.19% on CIFAR-10 and 5.52% on ImageNet-1K, compared to baselines. On Tiny-ImageNet, we obtained improved natural accuracy of 2.99\% on average and comparable adversarial accuracy. Our code and trained models are publicly available //github.com/xiaoyunxxy/MIMIR.
Recently, deep autoencoders have gained traction as a powerful method for implementing goal-oriented semantic communications systems. The idea is to train a mapping from the source domain directly to channel symbols, and vice versa. However, prior studies often focused on rate-distortion tradeoff and transmission delay, at the cost of increasing end-to-end complexity and thus latency. Moreover, the datasets used are often not reflective of real-world environments, and the results were not validated against real-world baseline systems, leading to an unfair comparison. In this paper, we study the problem of remote camera pose estimation and propose AdaSem, an adaptive semantic communications approach that optimizes the tradeoff between inference accuracy and end-to-end latency. We develop an adaptive semantic codec model, which encodes the source data into a dynamic number of symbols, based on the latent space distribution and the channel state feedback. We utilize a lightweight model for both transmitter and receiver to ensure comparable complexity to the baseline implemented in a real-world system. Extensive experiments on real-environment data show the effectiveness of our approach. When compared to a real implementation of a client-server camera relocalization service, AdaSem outperforms the baseline by reducing the end-to-end delay and estimation error by over 75% and 63%, respectively.
Unmanned Aerial Vehicles (UAVs) hold immense potential for critical applications, such as search and rescue operations, where accurate perception of indoor environments is paramount. However, the concurrent amalgamation of localization, 3D reconstruction, and semantic segmentation presents a notable hurdle, especially in the context of UAVs equipped with constrained power and computational resources. This paper presents a novel approach to address challenges in semantic information extraction and utilization within UAV operations. Our system integrates state-of-the-art visual SLAM to estimate a comprehensive 6-DoF pose and advanced object segmentation methods at the back end. To improve the computational and storage efficiency of the framework, we adopt a streamlined voxel-based 3D map representation - OctoMap to build a working system. Furthermore, the fusion algorithm is incorporated to obtain the semantic information of each frame from the front-end SLAM task, and the corresponding point. By leveraging semantic information, our framework enhances the UAV's ability to perceive and navigate through indoor spaces, addressing challenges in pose estimation accuracy and uncertainty reduction. Through Gazebo simulations, we validate the efficacy of our proposed system and successfully embed our approach into a Jetson Xavier AGX unit for real-world applications.
Bearing measurements,as the most common modality in nature, have recently gained traction in multi-robot systems to enhance mutual localization and swarm collaboration. Despite their advantages, challenges such as sensory noise, obstacle occlusion, and uncoordinated swarm motion persist in real-world scenarios, potentially leading to erroneous state estimation and undermining the system's flexibility, practicality, and robustness.In response to these challenges, in this paper we address theoretical and practical problem related to both mutual localization and swarm planning.Firstly, we propose a certifiable mutual localization algorithm.It features a concise problem formulation coupled with lossless convex relaxation, enabling independence from initial values and globally optimal relative pose recovery.Then, to explore how detection noise and swarm motion influence estimation optimality, we conduct a comprehensive analysis on the interplay between robots' mutual spatial relationship and mutual localization. We develop a differentiable metric correlated with swarm trajectories to explicitly evaluate the noise resistance of optimal estimation.By establishing a finite and pre-computable threshold for this metric and accordingly generating swarm trajectories, the estimation optimality can be strictly guaranteed under arbitrary noise. Based on these findings, an optimization-based swarm planner is proposed to generate safe and smooth trajectories, with consideration of both inter-robot visibility and estimation optimality.Through numerical simulations, we evaluate the optimality and certifiablity of our estimator, and underscore the significance of our planner in enhancing estimation performance.The results exhibit considerable potential of our methods to pave the way for advanced closed-loop intelligence in swarm systems.
The Tactile Internet paradigm is set to revolutionize human society by enabling skill-set delivery and haptic communication over ultra-reliable, low-latency networks. The emerging sixth-generation (6G) mobile communication systems are envisioned to underpin this Tactile Internet ecosystem at the network edge by providing ubiquitous global connectivity. However, apart from a multitude of opportunities of the Tactile Internet, security and privacy challenges emerge at the forefront. We believe that the recently standardized QUIC protocol, characterized by end-to-end encryption and reduced round-trip delay would serve as the backbone of Tactile Internet. In this article, we envision a futuristic scenario where a QUIC-enabled network uses the underlying 6G communication infrastructure to achieve the requirements for Tactile Internet. Interestingly this requires a deeper investigation of a wide range of security and privacy challenges in QUIC, that need to be mitigated for its adoption in Tactile Internet. Henceforth, this article reviews the existing security and privacy attacks in QUIC and their implication on users. Followed by that, we discuss state-of-the-art attack mitigation strategies and investigate some of their drawbacks with possible directions for future work
Owing to effective and flexible data acquisition, unmanned aerial vehicle (UAV) has recently become a hotspot across the fields of computer vision (CV) and remote sensing (RS). Inspired by recent success of deep learning (DL), many advanced object detection and tracking approaches have been widely applied to various UAV-related tasks, such as environmental monitoring, precision agriculture, traffic management. This paper provides a comprehensive survey on the research progress and prospects of DL-based UAV object detection and tracking methods. More specifically, we first outline the challenges, statistics of existing methods, and provide solutions from the perspectives of DL-based models in three research topics: object detection from the image, object detection from the video, and object tracking from the video. Open datasets related to UAV-dominated object detection and tracking are exhausted, and four benchmark datasets are employed for performance evaluation using some state-of-the-art methods. Finally, prospects and considerations for the future work are discussed and summarized. It is expected that this survey can facilitate those researchers who come from remote sensing field with an overview of DL-based UAV object detection and tracking methods, along with some thoughts on their further developments.
Most object recognition approaches predominantly focus on learning discriminative visual patterns while overlooking the holistic object structure. Though important, structure modeling usually requires significant manual annotations and therefore is labor-intensive. In this paper, we propose to "look into object" (explicitly yet intrinsically model the object structure) through incorporating self-supervisions into the traditional framework. We show the recognition backbone can be substantially enhanced for more robust representation learning, without any cost of extra annotation and inference speed. Specifically, we first propose an object-extent learning module for localizing the object according to the visual patterns shared among the instances in the same category. We then design a spatial context learning module for modeling the internal structures of the object, through predicting the relative positions within the extent. These two modules can be easily plugged into any backbone networks during training and detached at inference time. Extensive experiments show that our look-into-object approach (LIO) achieves large performance gain on a number of benchmarks, including generic object recognition (ImageNet) and fine-grained object recognition tasks (CUB, Cars, Aircraft). We also show that this learning paradigm is highly generalizable to other tasks such as object detection and segmentation (MS COCO). Project page: //github.com/JDAI-CV/LIO.
We propose a novel attention gate (AG) model for medical imaging that automatically learns to focus on target structures of varying shapes and sizes. Models trained with AGs implicitly learn to suppress irrelevant regions in an input image while highlighting salient features useful for a specific task. This enables us to eliminate the necessity of using explicit external tissue/organ localisation modules of cascaded convolutional neural networks (CNNs). AGs can be easily integrated into standard CNN architectures such as the U-Net model with minimal computational overhead while increasing the model sensitivity and prediction accuracy. The proposed Attention U-Net architecture is evaluated on two large CT abdominal datasets for multi-class image segmentation. Experimental results show that AGs consistently improve the prediction performance of U-Net across different datasets and training sizes while preserving computational efficiency. The code for the proposed architecture is publicly available.
注冊地址： 北京市海淀區羊坊店路18號2幢3層301-191