Potential radioactive hazards in full-dose positron emission tomography (PET) imaging remain a concern, whereas the quality of low-dose images is never desirable for clinical use. So it is of great interest to translate low-dose PET images into full-dose. Previous studies based on deep learning methods usually directly extract hierarchical features for reconstruction. We notice that the importance of each feature is different and they should be weighted dissimilarly so that tiny information can be captured by the neural network. Furthermore, the synthesis on some regions of interest is important in some applications. Here we propose a novel segmentation guided style-based generative adversarial network (SGSGAN) for PET synthesis. (1) We put forward a style-based generator employing style modulation, which specifically controls the hierarchical features in the translation process, to generate images with more realistic textures. (2) We adopt a task-driven strategy that couples a segmentation task with a generative adversarial network (GAN) framework to improve the translation performance. Extensive experiments show the superiority of our overall framework in PET synthesis, especially on those regions of interest.
Deep learning (DL) is gaining popularity as a parameter estimation method for quantitative MRI. A range of competing implementations have been proposed, relying on either supervised or self-supervised learning. Self-supervised approaches, sometimes referred to as unsupervised, have been loosely based on auto-encoders, whereas supervised methods have, to date, been trained on groundtruth labels. These two learning paradigms have been shown to have distinct strengths. Notably, self-supervised approaches have offered lower-bias parameter estimates than their supervised alternatives. This result is counterintuitive - incorporating prior knowledge with supervised labels should, in theory, lead to improved accuracy. In this work, we show that this apparent limitation of supervised approaches stems from the naive choice of groundtruth training labels. By training on labels which are deliberately not groundtruth, we show that the low-bias parameter estimation previously associated with self-supervised methods can be replicated - and improved on - within a supervised learning framework. This approach sets the stage for a single, unifying, deep learning parameter estimation framework, based on supervised learning, where trade-offs between bias and variance are made by careful adjustment of training label.
With the rapid development of deep learning, training Big Models (BMs) for multiple downstream tasks becomes a popular paradigm. Researchers have achieved various outcomes in the construction of BMs and the BM application in many fields. At present, there is a lack of research work that sorts out the overall progress of BMs and guides the follow-up research. In this paper, we cover not only the BM technologies themselves but also the prerequisites for BM training and applications with BMs, dividing the BM review into four parts: Resource, Models, Key Technologies and Application. We introduce 16 specific BM-related topics in those four parts, they are Data, Knowledge, Computing System, Parallel Training System, Language Model, Vision Model, Multi-modal Model, Theory&Interpretability, Commonsense Reasoning, Reliability&Security, Governance, Evaluation, Machine Translation, Text Generation, Dialogue and Protein Research. In each topic, we summarize clearly the current studies and propose some future research directions. At the end of this paper, we conclude the further development of BMs in a more general view.
Image animation brings life to the static object in the source image according to the driving video. Recent works attempt to perform motion transfer on arbitrary objects through unsupervised methods without using a priori knowledge. However, it remains a significant challenge for current unsupervised methods when there is a large pose gap between the objects in the source and driving images. In this paper, a new end-to-end unsupervised motion transfer framework is proposed to overcome such issue. Firstly, we propose thin-plate spline motion estimation to produce a more flexible optical flow, which warps the feature maps of the source image to the feature domain of the driving image. Secondly, in order to restore the missing regions more realistically, we leverage multi-resolution occlusion masks to achieve more effective feature fusion. Finally, additional auxiliary loss functions are designed to ensure that there is a clear division of labor in the network modules, encouraging the network to generate high-quality images. Our method can animate a variety of objects, including talking faces, human bodies, and pixel animations. Experiments demonstrate that our method performs better on most benchmarks than the state of the art with visible improvements in pose-related metrics.
Machine learning methods have revolutionized the discovery process of new molecules and materials. However, the intensive training process of neural networks for molecules with ever increasing complexity has resulted in exponential growth in computation cost, leading to long simulation time and high energy consumption. Photonic chip technology offers an alternative platform for implementing neural network with faster data processing and lower energy usage compared to digital computers. Here, we demonstrate the capability of photonic neural networks in predicting the quantum mechanical properties of molecules. Additionally, we show that multiple properties can be learned simultaneously in a photonic chip via a multi-task regression learning algorithm, which we believe is the first of its kind, as most previous works focus on implementing a network for the task of classification. Photonics technology are also naturally capable of implementing complex-valued neural networks at no additional hardware cost and we show that such neural networks outperform conventional real-valued networks for molecular property prediction. Our work opens the avenue for harnessing photonic technology for large-scale machine learning applications in molecular sciences such as drug discovery and materials design.
Anomaly detection (AD) has been an active research area in various domains. Yet, the increasing data scale, complexity, and dimension turn the traditional methods into challenging. Recently, the deep generative model, such as the variational autoencoder (VAE), has sparked a renewed interest in the AD problem. However, the probability distribution divergence used as the regularization is too strong, which causes the model cannot capture the manifold of the true data. In this paper, we propose the Projected Sliced Wasserstein (PSW) autoencoder-based anomaly detection method. Rooted in the optimal transportation, the PSW distance is a weaker distribution measure compared with $f$-divergence. In particular, the computation-friendly eigen-decomposition method is leveraged to find the principal component for slicing the high-dimensional data. In this case, the Wasserstein distance can be calculated with the closed-form, even the prior distribution is not Gaussian. Comprehensive experiments conducted on various real-world hyperspectral anomaly detection benchmarks demonstrate the superior performance of the proposed method.
News recommendation is an effective information dissemination solution in modern society. While recent years have witnessed many promising news recommendation models, they mostly capture the user-news interactions on the document-level in a static manner. However, in real-world scenarios, the news can be quite complex and diverse, blindly squeezing all the contents into an embedding vector can be less effective in extracting information compatible with the personalized preference of the users. In addition, user preferences in the news recommendation scenario can be highly dynamic, and a tailored dynamic mechanism should be designed for better recommendation performance. In this paper, we propose a novel dynamic news recommender model. For better understanding the news content, we leverage the attention mechanism to represent the news from the sentence-, element- and document-levels, respectively. For capturing users' dynamic preferences, the continuous time information is seamlessly incorporated into the computing of the attention weights. More specifically, we design a hierarchical attention network, where the lower layer learns the importance of different sentences and elements, and the upper layer captures the correlations between the previously interacted and the target news. To comprehensively model the dynamic characters, we firstly enhance the traditional attention mechanism by incorporating both absolute and relative time information, and then we propose a dynamic negative sampling method to optimize the users' implicit feedback. We conduct extensive experiments based on three real-world datasets to demonstrate our model's effectiveness. Our source code and pre-trained representations are available at https://github.com/lshowway/D-HAN.
Video Panoptic Segmentation (VPS) aims at assigning a class label to each pixel, uniquely segmenting and identifying all object instances consistently across all frames. Classic solutions usually decompose the VPS task into several sub-tasks and utilize multiple surrogates (e.g. boxes and masks, centres and offsets) to represent objects. However, this divide-and-conquer strategy requires complex post-processing in both spatial and temporal domains and is vulnerable to failures from surrogate tasks. In this paper, inspired by object-centric learning which learns compact and robust object representations, we present Slot-VPS, the first end-to-end framework for this task. We encode all panoptic entities in a video, including both foreground instances and background semantics, with a unified representation called panoptic slots. The coherent spatio-temporal object's information is retrieved and encoded into the panoptic slots by the proposed Video Panoptic Retriever, enabling it to localize, segment, differentiate, and associate objects in a unified manner. Finally, the output panoptic slots can be directly converted into the class, mask, and object ID of panoptic objects in the video. We conduct extensive ablation studies and demonstrate the effectiveness of our approach on two benchmark datasets, Cityscapes-VPS (\textit{val} and test sets) and VIPER (\textit{val} set), achieving new state-of-the-art performance of 63.7, 63.3 and 56.2 VPQ, respectively.
Action quality assessment (AQA) from videos is a challenging vision task since the relation between videos and action scores is difficult to model. Thus, action quality assessment has been widely studied in the literature. Traditionally, AQA task is treated as a regression problem to learn the underlying mappings between videos and action scores. More recently, the method of uncertainty score distribution learning (USDL) made success due to the introduction of label distribution learning (LDL). But USDL does not apply to dataset with continuous labels and needs a fixed variance in training. In this paper, to address the above problems, we further develop Distribution Auto-Encoder (DAE). DAE takes both advantages of regression algorithms and label distribution learning (LDL).Specifically, it encodes videos into distributions and uses the reparameterization trick in variational auto-encoders (VAE) to sample scores, which establishes a more accurate mapping between videos and scores. Meanwhile, a combined loss is constructed to accelerate the training of DAE. DAE-MT is further proposed to deal with AQA on multi-task datasets. We evaluate our DAE approach on MTL-AQA and JIGSAWS datasets. Experimental results on public datasets demonstrate that our method achieves state-of-the-arts under the Spearman's Rank Correlation: 0.9449 on MTL-AQA and 0.73 on JIGSAWS.
Purpose: To develop a semi-automated, AI-assisted workflow for segmentation of inflammatory lesions on STIR MRI of sacroiliac joints (SIJs) in adult patients with axial spondyloarthritis. Methods: Baseline human performance in manual segmentation of inflammatory lesions was first established in eight patients with axial spondyloarthritis recruited within a prospective study conducted between April 2018 and July 2019. To improve readers' consistency a semi-automated procedure was developed, comprising (1) manual segmentation of 'normal bone' and 'disease' regions (2) automatic segmentation of lesions, i.e., voxels in the disease region with outlying intensity with respect to the normal bone, and (3) human intervention to remove erroneously segmented areas. Segmentation of disease region (subchondral bone) was automated via supervised deep learning; 200 image slices (eight subjects) were used for algorithm training with cross validation, 48 (two subjects) - for testing and 500 (20 subjects) - for evaluation based on visual assessment. The data, code, and model are available at https://github.com/c-hepburn/Bone_MRI. Human and model performance were assessed in terms of Dice coefficient. Results: Intra-reader median Dice coefficients, evaluated from comparison of manual segmentation trials of inflammatory lesions, were 0.63 and 0.69 for the two readers, respectively. Inter-reader median Dice was in the range of 0.53 to 0.56 and increased to 0.84 using the semi-automated approach. Deep learning model ensemble showed average Dice of 0.94 in subchondral bone segmentation. Conclusions: We describe a semi-automated, AI-assisted workflow which improves the objectivity and consistency of radiological segmentation of inflammatory load in SIJs.