Non-autoregressive translation (NAT) model achieves a much faster inference speed than the autoregressive translation (AT) model because it can simultaneously predict all tokens during inference. However, its translation quality suffers from degradation compared to AT. And existing NAT methods only focus on improving the NAT model's performance but do not fully utilize it. In this paper, we propose a simple but effective method called "Candidate Soups," which can obtain high-quality translations while maintaining the inference speed of NAT models. Unlike previous approaches that pick the individual result and discard the remainders, Candidate Soups (CDS) can fully use the valuable information in the different candidate translations through model uncertainty. Extensive experiments on two benchmarks (WMT'14 EN-DE and WMT'16 EN-RO) demonstrate the effectiveness and generality of our proposed method, which can significantly improve the translation quality of various base models. More notably, our best variant outperforms the AT model on three translation tasks with 7.6 times speedup.
相關內容
Video object detection needs to solve feature degradation situations that rarely happen in the image domain. One solution is to use the temporal information and fuse the features from the neighboring frames. With Transformerbased object detectors getting a better performance on the image domain tasks, recent works began to extend those methods to video object detection. However, those existing Transformer-based video object detectors still follow the same pipeline as those used for classical object detectors, like enhancing the object feature representations by aggregation. In this work, we take a different perspective on video object detection. In detail, we improve the qualities of queries for the Transformer-based models by aggregation. To achieve this goal, we first propose a vanilla query aggregation module that weighted averages the queries according to the features of the neighboring frames. Then, we extend the vanilla module to a more practical version, which generates and aggregates queries according to the features of the input frames. Extensive experimental results validate the effectiveness of our proposed methods: On the challenging ImageNet VID benchmark, when integrated with our proposed modules, the current state-of-the-art Transformer-based object detectors can be improved by more than 2.4% on mAP and 4.2% on AP50.
While recent research advances in speaker diarization mostly focus on improving the quality of diarization results, there is also an increasing interest in improving the efficiency of diarization systems. In this paper, we demonstrate that a multi-stage clustering strategy that uses different clustering algorithms for input of different lengths can address multi-faceted challenges of on-device speaker diarization applications. Specifically, a fallback clusterer is used to handle short-form inputs; a main clusterer is used to handle medium-length inputs; and a pre-clusterer is used to compress long-form inputs before they are processed by the main clusterer. Both the main clusterer and the pre-clusterer can be configured with an upper bound of the computational complexity to adapt to devices with different resource constraints. This multi-stage clustering strategy is critical for streaming on-device speaker diarization systems, where the budgets of CPU, memory and battery are tight.
Conversational search applications offer the prospect of improved user experience in information seeking via agent support. However, it is not clear how searchers will respond to this mode of engagement, in comparison to a conventional user-driven search interface, such as those found in a standard web search engine. We describe a laboratory-based study directly comparing user behaviour for a conventional search interface (CSI) with that of an agent-mediated multiview conversational search interface (MCSI) which extends the CSI. User reaction and search outcomes of the two interfaces are compared using implicit evaluation using five analysis methods: claiming to have a better search experience in contrast to a corresponding standard search interface.
Safety is critical in robotic tasks. Energy function based methods have been introduced to address the problem. To ensure safety in the presence of control limits, we need to design an energy function that results in persistently feasible safe control at all system states. However, designing such an energy function for high-dimensional nonlinear systems remains challenging. Considering the fact that there are redundant dynamics in high dimensional systems with respect to the safety specifications, this paper proposes a novel approach called abstract safe control. We propose a system abstraction method that enables the design of energy functions on a low-dimensional model. Then we can synthesize the energy function with respect to the low-dimensional model to ensure persistent feasibility. The resulting safe controller can be directly transferred to other systems with the same abstraction, e.g., when a robot arm holds different tools. The proposed approach is demonstrated on a 7-DoF robot arm (14 states) both in simulation and real-world. Our method always finds feasible control and achieves zero safety violations in 500 trials on 5 different systems.
Body weight, as an essential physiological trait, is of considerable significance in many applications like body management, rehabilitation, and drug dosing for patient-specific treatments. Previous works on the body weight estimation task are mainly vision-based, using 2D/3D, depth, or infrared images, facing problems in illumination, occlusions, and especially privacy issues. The pressure mapping mattress is a non-invasive and privacy-preserving tool to obtain the pressure distribution image over the bed surface, which strongly correlates with the body weight of the lying person. To extract the body weight from this image, we propose a deep learning-based model, including a dual-branch network to extract the deep features and pose features respectively. A contrastive learning module is also combined with the deep-feature branch to help mine the mutual factors across different postures of every single subject. The two groups of features are then concatenated for the body weight regression task. To test the model's performance over different hardware and posture settings, we create a pressure image dataset of 10 subjects and 23 postures, using a self-made pressure-sensing bedsheet. This dataset, which is made public together with this paper, together with a public dataset, are used for the validation. The results show that our model outperforms the state-of-the-art algorithms over both 2 datasets. Our research constitutes an important step toward fully automatic weight estimation in both clinical and at-home practice. Our dataset is available for research purposes at: //github.com/USTCWzy/MassEstimation.
Video prediction is a complex time-series forecasting task with great potential in many use cases. However, conventional methods overemphasize accuracy while ignoring the slow prediction speed caused by complicated model structures that learn too much redundant information with excessive GPU memory consumption. Furthermore, conventional methods mostly predict frames sequentially (frame-by-frame) and thus are hard to accelerate. Consequently, valuable use cases such as real-time danger prediction and warning cannot achieve fast enough inference speed to be applicable in reality. Therefore, we propose a transformer-based keypoint prediction neural network (TKN), an unsupervised learning method that boost the prediction process via constrained information extraction and parallel prediction scheme. TKN is the first real-time video prediction solution to our best knowledge, while significantly reducing computation costs and maintaining other performance. Extensive experiments on KTH and Human3.6 datasets demonstrate that TKN predicts 11 times faster than existing methods while reducing memory consumption by 17.4% and achieving state-of-the-art prediction performance on average.
Despite considerable recent progress in Visual Question Answering (VQA) models, inconsistent or contradictory answers continue to cast doubt on their true reasoning capabilities. However, most proposed methods use indirect strategies or strong assumptions on pairs of questions and answers to enforce model consistency. Instead, we propose a novel strategy intended to improve model performance by directly reducing logical inconsistencies. To do this, we introduce a new consistency loss term that can be used by a wide range of the VQA models and which relies on knowing the logical relation between pairs of questions and answers. While such information is typically not available in VQA datasets, we propose to infer these logical relations using a dedicated language model and use these in our proposed consistency loss function. We conduct extensive experiments on the VQA Introspect and DME datasets and show that our method brings improvements to state-of-the-art VQA models, while being robust across different architectures and settings.
ODIN is an innovative approach that addresses the problem of dataset constraints by integrating generative AI models. Traditional zero-shot learning methods are constrained by the training dataset. To fundamentally overcome this limitation, ODIN attempts to mitigate the dataset constraints by generating on-demand datasets based on user requirements. ODIN consists of three main modules: a prompt generator, a text-to-image generator, and an image post-processor. To generate high-quality prompts and images, we adopted a large language model (e.g., ChatGPT), and a text-to-image diffusion model (e.g., Stable Diffusion), respectively. We evaluated ODIN on various datasets in terms of model accuracy and data diversity to demonstrate its potential, and conducted post-experiments for further investigation. Overall, ODIN is a feasible approach that enables Al to learn unseen knowledge beyond the training dataset.
In many visual systems, visual tracking often bases on RGB image sequences, in which some targets are invalid in low-light conditions, and tracking performance is thus affected significantly. Introducing other modalities such as depth and infrared data is an effective way to handle imaging limitations of individual sources, but multi-modal imaging platforms usually require elaborate designs and cannot be applied in many real-world applications at present. Near-infrared (NIR) imaging becomes an essential part of many surveillance cameras, whose imaging is switchable between RGB and NIR based on the light intensity. These two modalities are heterogeneous with very different visual properties and thus bring big challenges for visual tracking. However, existing works have not studied this challenging problem. In this work, we address the cross-modal object tracking problem and contribute a new video dataset, including 654 cross-modal image sequences with over 481K frames in total, and the average video length is more than 735 frames. To promote the research and development of cross-modal object tracking, we propose a new algorithm, which learns the modality-aware target representation to mitigate the appearance gap between RGB and NIR modalities in the tracking process. It is plug-and-play and could thus be flexibly embedded into different tracking frameworks. Extensive experiments on the dataset are conducted, and we demonstrate the effectiveness of the proposed algorithm in two representative tracking frameworks against 17 state-of-the-art tracking methods. We will release the dataset for free academic usage, dataset download link and code will be released soon.
Since hardware resources are limited, the objective of training deep learning models is typically to maximize accuracy subject to the time and memory constraints of training and inference. We study the impact of model size in this setting, focusing on Transformer models for NLP tasks that are limited by compute: self-supervised pretraining and high-resource machine translation. We first show that even though smaller Transformer models execute faster per iteration, wider and deeper models converge in significantly fewer steps. Moreover, this acceleration in convergence typically outpaces the additional computational overhead of using larger models. Therefore, the most compute-efficient training strategy is to counterintuitively train extremely large models but stop after a small number of iterations. This leads to an apparent trade-off between the training efficiency of large Transformer models and the inference efficiency of small Transformer models. However, we show that large models are more robust to compression techniques such as quantization and pruning than small models. Consequently, one can get the best of both worlds: heavily compressed, large models achieve higher accuracy than lightly compressed, small models.
小貼士
登錄享
相關主題
注冊地址： 北京市海淀區羊坊店路18號2幢3層301-191