Vision Transformers (ViTs) have emerged with superior performance on computer vision tasks compared to convolutional neural network (CNN)-based models. However, ViTs are mainly designed for image classification that generate single-scale low-resolution representations, which makes dense prediction tasks such as semantic segmentation challenging for ViTs. Therefore, we propose HRViT, which enhances ViTs to learn semantically-rich and spatially-precise multi-scale representations by integrating high-resolution multi-branch architectures with ViTs. We balance the model performance and efficiency of HRViT by various branch-block co-optimization techniques. Specifically, we explore heterogeneous branch designs, reduce the redundancy in linear layers, and augment the attention block with enhanced expressiveness. Those approaches enabled HRViT to push the Pareto frontier of performance and efficiency on semantic segmentation to a new level, as our evaluation results on ADE20K and Cityscapes show. HRViT achieves 50.20% mIoU on ADE20K and 83.16% mIoU on Cityscapes, surpassing state-of-the-art MiT and CSWin backbones with an average of +1.78 mIoU improvement, 28% parameter saving, and 21% FLOPs reduction, demonstrating the potential of HRViT as a strong vision backbone for semantic segmentation.
Federated learning (FL) enables distributed clients to learn a shared model for prediction while keeping the training data local on each client. However, existing FL requires fully-labeled data for training, which is inconvenient or sometimes infeasible to obtain due to the high labeling cost and the requirement of expertise. The lack of labels makes FL impractical in many realistic settings. Self-supervised learning can address this challenge by learning from unlabeled data such that FL can be widely used. Contrastive learning (CL), a self-supervised learning approach, can effectively learn data representations from unlabeled data. However, the distributed data collected on clients are usually not independent and identically distributed (non-IID) among clients, and each client may only have few classes of data, which degrades the performance of CL and learned representations. To tackle this problem, we propose a federated contrastive learning framework consisting of two approaches: feature fusion and neighborhood matching, by which a unified feature space among clients is learned for better data representations. Feature fusion provides remote features as accurate contrastive information to each client for better local learning. Neighborhood matching further aligns each client's local features to the remote features such that well-clustered features among clients can be learned. Extensive experiments show the effectiveness of the proposed framework. It outperforms other methods by 11\% on IID data and matches the performance of centralized learning.
We introduce PyTorchVideo, an open-source deep-learning library that provides a rich set of modular, efficient, and reproducible components for a variety of video understanding tasks, including classification, detection, self-supervised learning, and low-level processing. The library covers a full stack of video understanding tools including multimodal data loading, transformations, and models that reproduce state-of-the-art performance. PyTorchVideo further supports hardware acceleration that enables real-time inference on mobile devices. The library is based on PyTorch and can be used by any training framework; for example, PyTorchLightning, PySlowFast, or Classy Vision. PyTorchVideo is available at https://pytorchvideo.org/
Low-rank tensor compression has been proposed as a promising approach to reduce the memory and compute requirements of neural networks for their deployment on edge devices. Tensor compression reduces the number of parameters required to represent a neural network weight by assuming network weights possess a coarse higher-order structure. This coarse structure assumption has been applied to compress large neural networks such as VGG and ResNet. However modern state-of-the-art neural networks for computer vision tasks (i.e. MobileNet, EfficientNet) already assume a coarse factorized structure through depthwise separable convolutions, making pure tensor decomposition a less attractive approach. We propose to combine low-rank tensor decomposition with sparse pruning in order to take advantage of both coarse and fine structure for compression. We compress weights in SOTA architectures (MobileNetv3, EfficientNet, Vision Transformer) and compare this approach to sparse pruning and tensor decomposition alone.
Vision transformers (ViTs) have attracted much attention for their superior performance on computer vision tasks. To address their limitations of single-scale low-resolution representations, prior work adapts ViTs to high-resolution dense prediction tasks with hierarchical architectures to generate pyramid features. However, multi-scale representation learning is still under-explored on ViTs, given their classification-like sequential topology. To enhance ViTs with more capability to learn semantically-rich and spatially-precise multi-scale representations, in this work, we present an efficient integration of high-resolution multi-branch architectures with vision transformers, dubbed HRViT, pushing the Pareto front of dense prediction tasks to a new level. We explore heterogeneous branch design, reduce the redundancy in linear layers, and augment the model nonlinearity to balance the model performance and hardware efficiency. The proposed HRViT achieves 50.20% mIoU on ADE20K and 83.16% mIoU on Cityscapes for semantic segmentation tasks, surpassing state-of-the-art MiT and CSWin with an average of +1.78 mIoU improvement, 28% parameter reduction, and 21% FLOPs reduction, demonstrating the potential of HRViT as strong vision backbones.
The big data about music history contains information about time and users' behavior. Researchers could predict the trend of popular songs accurately by analyzing this data. The traditional trend prediction models can better predict the short trend than the long trend. In this paper, we proposed the improved LSTM Rolling Prediction Algorithm (LSTM-RPA), which combines LSTM historical input with current prediction results as model input for next time prediction. Meanwhile, this algorithm converts the long trend prediction task into multiple short trend prediction tasks. The evaluation results show that the LSTM-RPA model increased F score by 13.03%, 16.74%, 11.91%, 18.52%, compared with LSTM, BiLSTM, GRU and RNN. And our method outperforms tradi-tional sequence models, which are ARIMA and SMA, by 10.67% and 3.43% improvement in F score.Code: https://github.com/maliaosaide/lstm-rpa
From wearables to powerful smart devices, modern automatic speech recognition (ASR) models run on a variety of edge devices with different computational budgets. To navigate the Pareto front of model accuracy vs model size, researchers are trapped in a dilemma of optimizing model accuracy by training and fine-tuning models for each individual edge device while keeping the training GPU-hours tractable. In this paper, we propose Omni-sparsity DNN, where a single neural network can be pruned to generate optimized model for a large range of model sizes. We develop training strategies for Omni-sparsity DNN that allows it to find models along the Pareto front of word-error-rate (WER) vs model size while keeping the training GPU-hours to no more than that of training one singular model. We demonstrate the Omni-sparsity DNN with streaming E2E ASR models. Our results show great saving on training time and resources with similar or better accuracy on LibriSpeech compared to individually pruned sparse models: 2%-6.6% better WER on Test-other.
Automatic speech recognition (ASR) has become increasingly ubiquitous on modern edge devices. Past work developed streaming End-to-End (E2E) all-neural speech recognizers that can run compactly on edge devices. However, E2E ASR models are prone to overfitting and have difficulties in generalizing to unseen testing data. Various techniques have been proposed to regularize the training of ASR models, including layer normalization, dropout, spectrum data augmentation and speed distortions in the inputs. In this work, we present a simple yet effective noisy training strategy to further improve the E2E ASR model training. By introducing random noise to the parameter space during training, our method can produce smoother models at convergence that generalize better. We apply noisy training to improve both dense and sparse state-of-the-art Emformer models and observe consistent WER reduction. Specifically, when training Emformers with 90% sparsity, we achieve 12% and 14% WER improvements on the LibriSpeech Test-other and Test-clean data set, respectively.
Since the mapping relationship between definitized intra-interventional 2D X-ray and undefined pre-interventional 3D Computed Tomography(CT) is uncertain, auxiliary positioning devices or body markers, such as medical implants, are commonly used to determine this relationship. However, such approaches can not be widely used in clinical due to the complex realities. To determine the mapping relationship, and achieve a initializtion post estimation of human body without auxiliary equipment or markers, proposed method applies image segmentation and deep feature matching to directly match the 2D X-ray and 3D CT images. As a result, the well-trained network can directly predict the spatial correspondence between arbitrary 2D X-ray and 3D CT. The experimental results show that when combining our approach with the conventional approach, the achieved accuracy and speed can meet the basic clinical intervention needs, and it provides a new direction for intra-interventional registration.
Since the mapping relationship between definitized intra-interventional 2D X-ray and undefined pre-interventional 3D Computed Tomography(CT) is uncertain, auxiliary positioning devices or body markers, such as medical implants, are commonly used to determine this relationship. However, such approaches can not be widely used in clinical due to the complex realities. To determine the mapping relationship, and achieve a initializtion post estimation of human body without auxiliary equipment or markers, a cross-modal matching transformer network is proposed to matching 2D X-ray and 3D CT images directly. The proposed approach first deep learns skeletal features from 2D X-ray and 3D CT images. The features are then converted into 1D X-ray and CT representation vectors, which are combined using a multi-modal transformer. As a result, the well-trained network can directly predict the spatial correspondence between arbitrary 2D X-ray and 3D CT. The experimental results show that when combining our approach with the conventional approach, the achieved accuracy and speed can meet the basic clinical intervention needs, and it provides a new direction for intra-interventional registration.