PyTorchKR Papers

최신 논문

총 9,642편
태그·기간 등 상세 필터

카테고리 탐색

로그인 후 이 태그 구독

arXiv:2011.00927

Boost Image Captioning with Knowledge Reasoning

Automatically generating a human-like description for a given image is a potential research in artificial intelligence, which has attracted a great of attention recently. Most of the existing attention methods explore the mapping relationships between words in sentence and regions in image, such unpredictable matching manner sometimes causes inharmonious alignments that may reduce the quality of generated captions. In this paper, we make our efforts to reason about more accurate and meaningful captions. We first propose word attention to improve the correctness of visual attention when generating sequential descriptions word-by-word. The special word attention emphasizes on word importance when focusing on different regions of the input image, and makes full use of the internal annotation knowledge to assist the calculation of visual attention. Then, in order to reveal those incomprehensible intentions that cannot be expressed straightforwardly by machines, we introduce a new strategy to inject external knowledge extracted from knowledge graph into the encoder-decoder framework to facilitate meaningful captioning. Finally, we validate our model on two freely available captioning benchmarks: Microsoft COCO dataset and Flickr30k dataset. The results demonstrate that our approach achieves state-of-the-art performance and outperforms many of the existing approaches.

arXiv:2011.00168

Multimodal and self-supervised representation learning for automatic gesture recognition in surgical robotics

Self-supervised, multi-modal learning has been successful in holistic representation of complex scenarios. This can be useful to consolidate information from multiple modalities which have multiple, versatile uses. Its application in surgical robotics can lead to simultaneously developing a generalised machine understanding of the surgical process and reduce the dependency on quality, expert annotations which are generally difficult to obtain. We develop a self-supervised, multi-modal representation learning paradigm that learns representations for surgical gestures from video and kinematics. We use an encoder-decoder network configuration that encodes representations from surgical videos and decodes them to yield kinematics. We quantitatively demonstrate the efficacy of our learnt representations for gesture recognition (with accuracy between 69.6 % and 77.8 %), transfer learning across multiple tasks (with accuracy between 44.6 % and 64.8 %) and surgeon skill classification (with accuracy between 76.8 % and 81.2 %). Further, we qualitatively demonstrate that our self-supervised representations cluster in semantically meaningful properties (surgeon skill and gestures).

arXiv:2010.16117공식 코드 · Papers with Code

PyraPose: Feature Pyramids for Fast and Accurate Object Pose Estimation under Domain Shift

Object pose estimation enables robots to understand and interact with their environments. Training with synthetic data is necessary in order to adapt to novel situations. Unfortunately, pose estimation under domain shift, i.e., training on synthetic data and testing in the real world, is challenging. Deep learning-based approaches currently perform best when using encoder-decoder networks but typically do not generalize to new scenarios with different scene characteristics. We argue that patch-based approaches, instead of encoder-decoder networks, are more suited for synthetic-to-real transfer because local to global object information is better represented. To that end, we present a novel approach based on a specialized feature pyramid network to compute multi-scale features for creating pose hypotheses on different feature map resolutions in parallel. Our single-shot pose estimation approach is evaluated on multiple standard datasets and outperforms the state of the art by up to 35%. We also perform grasping experiments in the real world to demonstrate the advantage of using synthetic data to generalize to novel environments.

arXiv:2010.16056공식 코드 · Papers with Code

Generating Radiology Reports via Memory-driven Transformer

Medical imaging is frequently used in clinical practice and trials for diagnosis and treatment. Writing imaging reports is time-consuming and can be error-prone for inexperienced radiologists. Therefore, automatically generating radiology reports is highly desired to lighten the workload of radiologists and accordingly promote clinical automation, which is an essential task to apply artificial intelligence to the medical domain. In this paper, we propose to generate radiology reports with memory-driven Transformer, where a relational memory is designed to record key information of the generation process and a memory-driven conditional layer normalization is applied to incorporating the memory into the decoder of Transformer. Experimental results on two prevailing radiology report datasets, IU X-Ray and MIMIC-CXR, show that our proposed approach outperforms previous models with respect to both language generation metrics and clinical evaluations. Particularly, this is the first work reporting the generation results on MIMIC-CXR to the best of our knowledge. Further analyses also demonstrate that our approach is able to generate long reports with necessary medical terms as well as meaningful image-text attention mappings.

arXiv:2010.15881공식 코드 · Papers with Code

Less is More: Data-Efficient Complex Question Answering over Knowledge Bases

Question answering is an effective method for obtaining information from knowledge bases (KB). In this paper, we propose the Neural-Symbolic Complex Question Answering (NS-CQA) model, a data-efficient reinforcement learning framework for complex question answering by using only a modest number of training samples. Our framework consists of a neural generator and a symbolic executor that, respectively, transforms a natural-language question into a sequence of primitive actions, and executes them over the knowledge base to compute the answer. We carefully formulate a set of primitive symbolic actions that allows us to not only simplify our neural network design but also accelerate model convergence. To reduce search space, we employ the copy and masking mechanisms in our encoder-decoder architecture to drastically reduce the decoder output vocabulary and improve model generalizability. We equip our model with a memory buffer that stores high-reward promising programs. Besides, we propose an adaptive reward function. By comparing the generated trial with the trials stored in the memory buffer, we derive the curriculum-guided reward bonus, i.e., the proximity and the novelty. To mitigate the sparse reward problem, we combine the adaptive reward and the reward bonus, reshaping the sparse reward into dense feedback. Also, we encourage the model to generate new trials to avoid imitating the spurious trials while making the model remember the past high-reward trials to improve data efficiency. Our NS-CQA model is evaluated on two datasets: CQA, a recent large-scale complex question answering dataset, and WebQuestionsSP, a multi-hop question answering dataset. On both datasets, our model outperforms the state-of-the-art models. Notably, on CQA, NS-CQA performs well on questions with higher complexity, while only using approximately 1% of the total training samples.

arXiv:2010.15560

Genetic U-Net: Automatically Designed Deep Networks for Retinal Vessel Segmentation Using a Genetic Algorithm

Recently, many methods based on hand-designed convolutional neural networks (CNNs) have achieved promising results in automatic retinal vessel segmentation. However, these CNNs remain constrained in capturing retinal vessels in complex fundus images. To improve their segmentation performance, these CNNs tend to have many parameters, which may lead to overfitting and high computational complexity. Moreover, the manual design of competitive CNNs is time-consuming and requires extensive empirical knowledge. Herein, a novel automated design method, called Genetic U-Net, is proposed to generate a U-shaped CNN that can achieve better retinal vessel segmentation but with fewer architecture-based parameters, thereby addressing the above issues. First, we devised a condensed but flexible search space based on a U-shaped encoder-decoder. Then, we used an improved genetic algorithm to identify better-performing architectures in the search space and investigated the possibility of finding a superior network architecture with fewer parameters. The experimental results show that the architecture obtained using the proposed method offered a superior performance with less than 1% of the number of the original U-Net parameters in particular and with significantly fewer parameters than other state-of-the-art models. Furthermore, through in-depth investigation of the experimental results, several effective operations and patterns of networks to generate superior retinal vessel segmentations were identified.

arXiv:2010.15831공식 코드 · Papers with Code

RelationNet++: Bridging Visual Representations for Object Detection via Transformer Decoder

Existing object detection frameworks are usually built on a single format of object/part representation, i.e., anchor/proposal rectangle boxes in RetinaNet and Faster R-CNN, center points in FCOS and RepPoints, and corner points in CornerNet. While these different representations usually drive the frameworks to perform well in different aspects, e.g., better classification or finer localization, it is in general difficult to combine these representations in a single framework to make good use of each strength, due to the heterogeneous or non-grid feature extraction by different representations. This paper presents an attention-based decoder module similar as that in Transformer~\cite{vaswani2017attention} to bridge other representations into a typical object detector built on a single representation format, in an end-to-end fashion. The other representations act as a set of \emph{key} instances to strengthen the main \emph{query} representation features in the vanilla detectors. Novel techniques are proposed towards efficient computation of the decoder module, including a \emph{key sampling} approach and a \emph{shared location embedding} approach. The proposed module is named \emph{bridging visual representations} (BVR). It can perform in-place and we demonstrate its broad effectiveness in bridging other representations into prevalent object detection frameworks, including RetinaNet, Faster R-CNN, FCOS and ATSS, where about $1.5\sim3.0$ AP improvements are achieved. In particular, we improve a state-of-the-art framework with a strong backbone by about $2.0$ AP, reaching $52.7$ AP on COCO test-dev. The resulting network is named RelationNet++. The code will be available at https://github.com/microsoft/RelationNet2.

arXiv:2010.15643

Free-Form Image Inpainting via Contrastive Attention Network

Most deep learning based image inpainting approaches adopt autoencoder or its variants to fill missing regions in images. Encoders are usually utilized to learn powerful representational spaces, which are important for dealing with sophisticated learning tasks. Specifically, in image inpainting tasks, masks with any shapes can appear anywhere in images (i.e., free-form masks) which form complex patterns. It is difficult for encoders to capture such powerful representations under this complex situation. To tackle this problem, we propose a self-supervised Siamese inference network to improve the robustness and generalization. It can encode contextual semantics from full resolution images and obtain more discriminative representations. we further propose a multi-scale decoder with a novel dual attention fusion module (DAF), which can combine both the restored and known regions in a smooth way. This multi-scale architecture is beneficial for decoding discriminative representations learned by encoders into images layer by layer. In this way, unknown regions will be filled naturally from outside to inside. Qualitative and quantitative experiments on multiple datasets, including facial and natural datasets (i.e., Celeb-HQ, Pairs Street View, Places2 and ImageNet), demonstrate that our proposed method outperforms state-of-the-art methods in generating high-quality inpainting results.

arXiv:2010.14685

Parameterized Neural Ordinary Differential Equations: Applications to Computational Physics Problems

This work proposes an extension of neural ordinary differential equations (NODEs) by introducing an additional set of ODE input parameters to NODEs. This extension allows NODEs to learn multiple dynamics specified by the input parameter instances. Our extension is inspired by the concept of parameterized ordinary differential equations, which are widely investigated in computational science and engineering contexts, where characteristics of the governing equations vary over the input parameters. We apply the proposed parameterized NODEs (PNODEs) for learning latent dynamics of complex dynamical processes that arise in computational physics, which is an essential component for enabling rapid numerical simulations for time-critical physics applications. For this, we propose an encoder-decoder-type framework, which models latent dynamics as PNODEs. We demonstrate the effectiveness of PNODEs with important benchmark problems from computational physics.

arXiv:2010.15025

Non-Autoregressive Transformer ASR with CTC-Enhanced Decoder Input

Non-autoregressive (NAR) transformer models have achieved significantly inference speedup but at the cost of inferior accuracy compared to autoregressive (AR) models in automatic speech recognition (ASR). Most of the NAR transformers take a fixed-length sequence filled with MASK tokens or a redundant sequence copied from encoder states as decoder input, they cannot provide efficient target-side information thus leading to accuracy degradation. To address this problem, we propose a CTC-enhanced NAR transformer, which generates target sequence by refining predictions of the CTC module. Experimental results show that our method outperforms all previous NAR counterparts and achieves 50x faster decoding speed than a strong AR baseline with only 0.0 ~ 0.3 absolute CER degradation on Aishell-1 and Aishell-2 datasets.

arXiv:2010.14920

Bridging the Modality Gap for Speech-to-Text Translation

End-to-end speech translation aims to translate speech in one language into text in another language via an end-to-end way. Most existing methods employ an encoder-decoder structure with a single encoder to learn acoustic representation and semantic information simultaneously, which ignores the speech-and-text modality differences and makes the encoder overloaded, leading to great difficulty in learning such a model. To address these issues, we propose a Speech-to-Text Adaptation for Speech Translation (STAST) model which aims to improve the end-to-end model performance by bridging the modality gap between speech and text. Specifically, we decouple the speech translation encoder into three parts and introduce a shrink mechanism to match the length of speech representation with that of the corresponding text transcription. To obtain better semantic representation, we completely integrate a text-based translation model into the STAST so that two tasks can be trained in the same latent space. Furthermore, we introduce a cross-modal adaptation method to close the distance between speech and text representation. Experimental results on English-French and English-German speech translation corpora have shown that our model significantly outperforms strong baselines, and achieves the new state-of-the-art performance.

arXiv:2010.14794공식 코드 · Papers with Code

Seen and Unseen emotional style transfer for voice conversion with a new emotional speech dataset

Emotional voice conversion aims to transform emotional prosody in speech while preserving the linguistic content and speaker identity. Prior studies show that it is possible to disentangle emotional prosody using an encoder-decoder network conditioned on discrete representation, such as one-hot emotion labels. Such networks learn to remember a fixed set of emotional styles. In this paper, we propose a novel framework based on variational auto-encoding Wasserstein generative adversarial network (VAW-GAN), which makes use of a pre-trained speech emotion recognition (SER) model to transfer emotional style during training and at run-time inference. In this way, the network is able to transfer both seen and unseen emotional style to a new utterance. We show that the proposed framework achieves remarkable performance by consistently outperforming the baseline framework. This paper also marks the release of an emotional speech dataset (ESD) for voice conversion, which has multiple speakers and languages.

arXiv:2010.14725

CASS-NAT: CTC Alignment-based Single Step Non-autoregressive Transformer for Speech Recognition

We propose a CTC alignment-based single step non-autoregressive transformer (CASS-NAT) for speech recognition. Specifically, the CTC alignment contains the information of (a) the number of tokens for decoder input, and (b) the time span of acoustics for each token. The information are used to extract acoustic representation for each token in parallel, referred to as token-level acoustic embedding which substitutes the word embedding in autoregressive transformer (AT) to achieve parallel generation in decoder. During inference, an error-based alignment sampling method is proposed to be applied to the CTC output space, reducing the WER and retaining the parallelism as well. Experimental results show that the proposed method achieves WERs of 3.8%/9.1% on Librispeech test clean/other dataset without an external LM, and a CER of 5.8% on Aishell1 Mandarin corpus, respectively1. Compared to the AT baseline, the CASS-NAT has a performance reduction on WER, but is 51.2x faster in terms of RTF. When decoding with an oracle CTC alignment, the lower bound of WER without LM reaches 2.3% on the test-clean set, indicating the potential of the proposed method.

arXiv:2010.14606

Cascaded encoders for unifying streaming and non-streaming ASR

End-to-end (E2E) automatic speech recognition (ASR) models, by now, have shown competitive performance on several benchmarks. These models are structured to either operate in streaming or non-streaming mode. This work presents cascaded encoders for building a single E2E ASR model that can operate in both these modes simultaneously. The proposed model consists of streaming and non-streaming encoders. Input features are first processed by the streaming encoder; the non-streaming encoder operates exclusively on the output of the streaming encoder. A single decoder then learns to decode either using the output of the streaming or the non-streaming encoder. Results show that this model achieves similar word error rates (WER) as a standalone streaming model when operating in streaming mode, and obtains 10% -- 27% relative improvement when operating in non-streaming mode. Our results also show that the proposed approach outperforms existing E2E two-pass models, especially on long-form speech.

arXiv:2010.14649공식 코드 · Papers with Code

Learning Contextualised Cross-lingual Word Embeddings and Alignments for Extremely Low-Resource Languages Using Parallel Corpora

We propose a new approach for learning contextualised cross-lingual word embeddings based on a small parallel corpus (e.g. a few hundred sentence pairs). Our method obtains word embeddings via an LSTM encoder-decoder model that simultaneously translates and reconstructs an input sentence. Through sharing model parameters among different languages, our model jointly trains the word embeddings in a common cross-lingual space. We also propose to combine word and subword embeddings to make use of orthographic similarities across different languages. We base our experiments on real-world data from endangered languages, namely Yongning Na, Shipibo-Konibo, and Griko. Our experiments on bilingual lexicon induction and word alignment tasks show that our model outperforms existing methods by a large margin for most language pairs. These results demonstrate that, contrary to common belief, an encoder-decoder translation model is beneficial for learning cross-lingual representations even in extremely low-resource conditions. Furthermore, our model also works well on high-resource conditions, achieving state-of-the-art performance on a German-English word-alignment task.

arXiv:2010.14442

Generating 3D Molecular Structures Conditional on a Receptor Binding Site with Deep Generative Models

Deep generative models have been applied with increasing success to the generation of two dimensional molecules as SMILES strings and molecular graphs. In this work we describe for the first time a deep generative model that can generate 3D molecular structures conditioned on a three-dimensional (3D) binding pocket. Using convolutional neural networks, we encode atomic density grids into separate receptor and ligand latent spaces. The ligand latent space is variational to support sampling of new molecules. A decoder network generates atomic densities of novel ligands conditioned on the receptor. Discrete atoms are then fit to these continuous densities to create molecular structures. We show that valid and unique molecules can be readily sampled from the variational latent space defined by a reference `seed' structure and generated structures have reasonable interactions with the binding site. As structures are sampled farther in latent space from the seed structure, the novelty of the generated structures increases, but the predicted binding affinity decreases. Overall, we demonstrate the feasibility of conditional 3D molecular structure generation and provide a starting point for methods that also explicitly optimize for desired molecular properties, such as high binding affinity.

arXiv:2010.12670

3DBooSTeR: 3D Body Shape and Texture Recovery

We propose 3DBooSTeR, a novel method to recover a textured 3D body mesh from a textured partial 3D scan. With the advent of virtual and augmented reality, there is a demand for creating realistic and high-fidelity digital 3D human representations. However, 3D scanning systems can only capture the 3D human body shape up to some level of defects due to its complexity, including occlusion between body parts, varying levels of details, shape deformations and the articulated skeleton. Textured 3D mesh completion is thus important to enhance 3D acquisitions. The proposed approach decouples the shape and texture completion into two sequential tasks. The shape is recovered by an encoder-decoder network deforming a template body mesh. The texture is subsequently obtained by projecting the partial texture onto the template mesh before inpainting the corresponding texture map with a novel approach. The approach is validated on the 3DBodyTex.v2 dataset.

arXiv:2010.14233

Align-Refine: Non-Autoregressive Speech Recognition via Iterative Realignment

Non-autoregressive models greatly improve decoding speed over typical sequence-to-sequence models, but suffer from degraded performance. Infilling and iterative refinement models make up some of this gap by editing the outputs of a non-autoregressive model, but are constrained in the edits that they can make. We propose iterative realignment, where refinements occur over latent alignments rather than output sequence space. We demonstrate this in speech recognition with Align-Refine, an end-to-end Transformer-based model which refines connectionist temporal classification (CTC) alignments to allow length-changing insertions and deletions. Align-Refine outperforms Imputer and Mask-CTC, matching an autoregressive baseline on WSJ at 1/14th the real-time factor and attaining a LibriSpeech test-other WER of 9.0% without an LM. Our model is strong even in one iteration with a shallower decoder.

arXiv:2010.14061공식 코드 · Papers with Code

Jointly Optimizing State Operation Prediction and Value Generation for Dialogue State Tracking

We investigate the problem of multi-domain Dialogue State Tracking (DST) with open vocabulary. Existing approaches exploit BERT encoder and copy-based RNN decoder, where the encoder predicts the state operation, and the decoder generates new slot values. However, in such a stacked encoder-decoder structure, the operation prediction objective only affects the BERT encoder and the value generation objective mainly affects the RNN decoder. In this paper, we propose a purely Transformer-based framework, where a single BERT works as both the encoder and the decoder. In so doing, the operation prediction objective and the value generation objective can jointly optimize this BERT for DST. At the decoding step, we re-use the hidden states of the encoder in the self-attention mechanism of the corresponding decoder layers to construct a flat encoder-decoder architecture for effective parameter updating. Experimental results show that our approach substantially outperforms the existing state-of-the-art framework, and it also achieves very competitive performance to the best ontology-based approaches.

arXiv:2010.12876

Electromagnetic Source Imaging via a Data-Synthesis-Based Convolutional Encoder-Decoder Network

Electromagnetic source imaging (ESI) requires solving a highly ill-posed inverse problem. To seek a unique solution, traditional ESI methods impose various forms of priors that may not accurately reflect the actual source properties, which may hinder their broad applications. To overcome this limitation, in this paper a novel data-synthesized spatio-temporally convolutional encoder-decoder network method termed DST-CedNet is proposed for ESI. DST-CedNet recasts ESI as a machine learning problem, where discriminative learning and latent-space representations are integrated in a convolutional encoder-decoder network (CedNet) to learn a robust mapping from the measured electroencephalography/magnetoencephalography (E/MEG) signals to the brain activity. In particular, by incorporating prior knowledge regarding dynamical brain activities, a novel data synthesis strategy is devised to generate large-scale samples for effectively training CedNet. This stands in contrast to traditional ESI methods where the prior information is often enforced via constraints primarily aimed for mathematical convenience. Extensive numerical experiments as well as analysis of a real MEG and Epilepsy EEG dataset demonstrate that DST-CedNet outperforms several state-of-the-art ESI methods in robustly estimating source signals under a variety of source configurations.