PyTorchKR Papers

최신 논문

총 9,642편
태그·기간 등 상세 필터

카테고리 탐색

로그인 후 이 태그 구독

arXiv:2110.01659

Cross-Modal Virtual Sensing for Combustion Instability Monitoring

In many cyber-physical systems, imaging can be an important but expensive or 'difficult to deploy' sensing modality. One such example is detecting combustion instability using flame images, where deep learning frameworks have demonstrated state-of-the-art performance. The proposed frameworks are also shown to be quite trustworthy such that domain experts can have sufficient confidence to use these models in real systems to prevent unwanted incidents. However, flame imaging is not a common sensing modality in engine combustors today. Therefore, the current roadblock exists on the hardware side regarding the acquisition and processing of high-volume flame images. On the other hand, the acoustic pressure time series is a more feasible modality for data collection in real combustors. To utilize acoustic time series as a sensing modality, we propose a novel cross-modal encoder-decoder architecture that can reconstruct cross-modal visual features from acoustic pressure time series in combustion systems. With the "distillation" of cross-modal features, the results demonstrate that the detection accuracy can be enhanced using the virtual visual sensing modality. By providing the benefit of cross-modal reconstruction, our framework can prove to be useful in different domains well beyond the power generation and transportation industries.

arXiv:2110.02067

Teach Me What to Say and I Will Learn What to Pick: Unsupervised Knowledge Selection Through Response Generation with Pretrained Generative Models

Knowledge Grounded Conversation Models (KGCM) are usually based on a selection/retrieval module and a generation module, trained separately or simultaneously, with or without having access to a gold knowledge option. With the introduction of large pre-trained generative models, the selection and generation part have become more and more entangled, shifting the focus towards enhancing knowledge incorporation (from multiple sources) instead of trying to pick the best knowledge option. These approaches however depend on knowledge labels and/or a separate dense retriever for their best performance. In this work we study the unsupervised selection abilities of pre-trained generative models (e.g. BART) and show that by adding a score-and-aggregate module between encoder and decoder, they are capable of learning to pick the proper knowledge through minimising the language modelling loss (i.e. without having access to knowledge labels). Trained as such, our model - K-Mine - shows competitive selection and generation performance against models that benefit from knowledge labels and/or separate dense retriever.

arXiv:2110.01939공식 코드 · Papers with Code

Double Encoder-Decoder Networks for Gastrointestinal Polyp Segmentation

Polyps represent an early sign of the development of Colorectal Cancer. The standard procedure for their detection consists of colonoscopic examination of the gastrointestinal tract. However, the wide range of polyp shapes and visual appearances, as well as the reduced quality of this image modality, turn their automatic identification and segmentation with computational tools into a challenging computer vision task. In this work, we present a new strategy for the delineation of gastrointestinal polyps from endoscopic images based on a direct extension of common encoder-decoder networks for semantic segmentation. In our approach, two pretrained encoder-decoder networks are sequentially stacked: the second network takes as input the concatenation of the original frame and the initial prediction generated by the first network, which acts as an attention mechanism enabling the second network to focus on interesting areas within the image, thereby improving the quality of its predictions. Quantitative evaluation carried out on several polyp segmentation databases shows that double encoder-decoder networks clearly outperform their single encoder-decoder counterparts in all cases. In addition, our best double encoder-decoder combination attains excellent segmentation accuracy and reaches state-of-the-art performance results in all the considered datasets, with a remarkable boost of accuracy on images extracted from datasets not used for training.

arXiv:2110.01811공식 코드 · Papers with Code

On the Complementarity between Pre-Training and Back-Translation for Neural Machine Translation

Pre-training (PT) and back-translation (BT) are two simple and powerful methods to utilize monolingual data for improving the model performance of neural machine translation (NMT). This paper takes the first step to investigate the complementarity between PT and BT. We introduce two probing tasks for PT and BT respectively and find that PT mainly contributes to the encoder module while BT brings more benefits to the decoder. Experimental results show that PT and BT are nicely complementary to each other, establishing state-of-the-art performances on the WMT16 English-Romanian and English-Russian benchmarks. Through extensive analyses on sentence originality and word frequency, we also demonstrate that combining Tagged BT with PT is more helpful to their complementarity, leading to better translation quality. Source code is freely available at https://github.com/SunbowLiu/PTvsBT.

arXiv:2110.01164

Decoupling Speaker-Independent Emotions for Voice Conversion Via Source-Filter Networks

Emotional voice conversion (VC) aims to convert a neutral voice to an emotional (e.g. happy) one while retaining the linguistic information and speaker identity. We note that the decoupling of emotional features from other speech information (such as speaker, content, etc.) is the key to achieving remarkable performance. Some recent attempts about speech representation decoupling on the neutral speech can not work well on the emotional speech, due to the more complex acoustic properties involved in the latter. To address this problem, here we propose a novel Source-Filter-based Emotional VC model (SFEVC) to achieve proper filtering of speaker-independent emotion features from both the timbre and pitch features. Our SFEVC model consists of multi-channel encoders, emotion separate encoders, and one decoder. Note that all encoder modules adopt a designed information bottlenecks auto-encoder. Additionally, to further improve the conversion quality for various emotions, a novel two-stage training strategy based on the 2D Valence-Arousal (VA) space was proposed. Experimental results show that the proposed SFEVC along with a two-stage training strategy outperforms all baselines and achieves the state-of-the-art performance in speaker-independent emotional VC with nonparallel data.

arXiv:2110.01014

EAR-U-Net: EfficientNet and attention-based residual U-Net for automatic liver segmentation in CT

Purpose: This paper proposes a new network framework called EAR-U-Net, which leverages EfficientNetB4, attention gate, and residual learning techniques to achieve automatic and accurate liver segmentation. Methods: The proposed method is based on the U-Net framework. First, we use EfficientNetB4 as the encoder to extract more feature information during the encoding stage. Then, an attention gate is introduced in the skip connection to eliminate irrelevant regions and highlight features of a specific segmentation task. Finally, to alleviate the problem of gradient vanishment, we replace the traditional convolution of the decoder with a residual block to improve the segmentation accuracy. Results: We verified the proposed method on the LiTS17 and SLiver07 datasets and compared it with classical networks such as FCN, U-Net, Attention U-Net, and Attention Res-U-Net. In the Sliver07 evaluation, the proposed method achieved the best segmentation performance on all five standard metrics. Meanwhile, in the LiTS17 assessment, the best performance is obtained except for a slight inferior on RVD. Moreover, we also participated in the MICCIA-LiTS17 challenge, and the Dice per case score was 0.952. Conclusion: The proposed method's qualitative and quantitative results demonstrated its applicability in liver segmentation and proved its good prospect in computer-assisted liver segmentation.

arXiv:2110.00896

Disarranged Zone Learning (DZL): An unsupervised and dynamic automatic stenosis recognition methodology based on coronary angiography

We proposed a novel unsupervised methodology named Disarranged Zone Learning (DZL) to automatically recognize stenosis in coronary angiography. The methodology firstly disarranges the frames in a video, secondly it generates an effective zone and lastly trains an encoder-decoder GRU model to learn the capability to recover disarranged frames. The breakthrough of our study is to discover and validate the Sequence Intensity (Recover Difficulty) is a measure of Coronary Artery Stenosis Status. Hence, the prediction accuracy of DZL is used as an approximator of coronary stenosis indicator. DZL is an unsupervised methodology and no label engineering effort is needed, the sub GRU model in DZL works as a self-supervised approach. So DZL could theoretically utilize infinitely huge amounts of coronary angiographies to learn and improve performance without laborious data labeling. There is no data preprocessing precondition to run DZL as it dynamically utilizes the whole video, hence it is easy to be implemented and generalized to overcome the data heterogeneity of coronary angiography. The overall average precision score achieves 0.93, AUC achieves 0.8 for this pure methodology. The highest segmented average precision score is 0.98 and the highest segmented AUC is 0.87 for coronary occlusion indicator. Finally, we developed a software demo to implement DZL methodology.

arXiv:2110.00682

Multi-view SA-LA Net: A framework for simultaneous segmentation of RV on multi-view cardiac MR Images

We proposed a multi-view SA-LA model for simultaneous segmentation of RV on the short-axis (SA) and long-axis (LA) cardiac MR images. The multi-view SA-LA model is a multi-encoder, multi-decoder U-Net architecture based on the U-Net model. One encoder-decoder pair segments the RV on SA images and the other pair on LA images. Multi-view SA-LA model assembles an extremely rich set of synergistic features, at the root of the encoder branch, by combining feature maps learned from matched SA and LA cardiac MR images. Segmentation performance is further enhanced by: (1) incorporating spatial context of LV as a prior and (2) performing deep supervision in the last three layers of the decoder branch. Multi-view SA-LA model was extensively evaluated on the MICCAI 2021 Multi- Disease, Multi-View, and Multi- Centre RV Segmentation Challenge dataset (M&Ms-2021). M&Ms-2021 dataset consists of multi-phase, multi-view cardiac MR images of 360 subjects acquired at four clinical centers with three different vendors. On the challenge cohort (160 subjects), the proposed multi-view SA-LA model achieved a Dice Score of 91% and Hausdorff distance of 11.2 mm on short-axis images and a Dice Score of 89.6% and Hausdorff distance of 8.1 mm on long-axis images. Moreover, multi-view SA-LA model exhibited strong generalization to unseen RV related pathologies including Dilated Right Ventricle (DSC: SA 91.41%, LA 89.63%) and Tricuspidal Regurgitation (DSC: SA 91.40%, LA 90.40%) with low variance (std_DSC: SA <5%, LA<6%).

arXiv:2110.00335공식 코드 · Papers with Code

Geometry Attention Transformer with Position-aware LSTMs for Image Captioning

In recent years, transformer structures have been widely applied in image captioning with impressive performance. For good captioning results, the geometry and position relations of different visual objects are often thought of as crucial information. Aiming to further promote image captioning by transformers, this paper proposes an improved Geometry Attention Transformer (GAT) model. In order to further leverage geometric information, two novel geometry-aware architectures are designed respectively for the encoder and decoder in our GAT. Besides, this model includes the two work modules: 1) a geometry gate-controlled self-attention refiner, for explicitly incorporating relative spatial information into image region representations in encoding steps, and 2) a group of position-LSTMs, for precisely informing the decoder of relative word position in generating caption texts. The experiment comparisons on the datasets MS COCO and Flickr30K show that our GAT is efficient, and it could often outperform current state-of-the-art image captioning models.

arXiv:2109.14927공식 코드 · Papers with Code

BERT got a Date: Introducing Transformers to Temporal Tagging

Temporal expressions in text play a significant role in language understanding and correctly identifying them is fundamental to various retrieval and natural language processing systems. Previous works have slowly shifted from rule-based to neural architectures, capable of tagging expressions with higher accuracy. However, neural models can not yet distinguish between different expression types at the same level as their rule-based counterparts. In this work, we aim to identify the most suitable transformer architecture for joint temporal tagging and type classification, as well as, investigating the effect of semi-supervised training on the performance of these systems. Based on our study of token classification variants and encoder-decoder architectures, we present a transformer encoder-decoder model using the RoBERTa language model as our best performing system. By supplementing training resources with weakly labeled data from rule-based systems, our model surpasses previous works in temporal tagging and type classification, especially on rare classes. Our code and pre-trained experiments are available at: https://github.com/satya77/Transformer_Temporal_Tagger

arXiv:2109.15142공식 코드 · Papers with Code

Redesigning the Transformer Architecture with Insights from Multi-particle Dynamical Systems

The Transformer and its variants have been proven to be efficient sequence learners in many different domains. Despite their staggering success, a critical issue has been the enormous number of parameters that must be trained (ranging from $10^7$ to $10^{11}$) along with the quadratic complexity of dot-product attention. In this work, we investigate the problem of approximating the two central components of the Transformer -- multi-head self-attention and point-wise feed-forward transformation, with reduced parameter space and computational complexity. We build upon recent developments in analyzing deep neural networks as numerical solvers of ordinary differential equations. Taking advantage of an analogy between Transformer stages and the evolution of a dynamical system of multiple interacting particles, we formulate a temporal evolution scheme, TransEvolve, to bypass costly dot-product attention over multiple stacked layers. We perform exhaustive experiments with TransEvolve on well-known encoder-decoder as well as encoder-only tasks. We observe that the degree of approximation (or inversely, the degree of parameter reduction) has different effects on the performance, depending on the task. While in the encoder-decoder regime, TransEvolve delivers performances comparable to the original Transformer, in encoder-only tasks it consistently outperforms Transformer along with several subsequent variants.

arXiv:2109.15047공식 코드 · Papers with Code

Deep Contextual Video Compression

Most of the existing neural video compression methods adopt the predictive coding framework, which first generates the predicted frame and then encodes its residue with the current frame. However, as for compression ratio, predictive coding is only a sub-optimal solution as it uses simple subtraction operation to remove the redundancy across frames. In this paper, we propose a deep contextual video compression framework to enable a paradigm shift from predictive coding to conditional coding. In particular, we try to answer the following questions: how to define, use, and learn condition under a deep video compression framework. To tap the potential of conditional coding, we propose using feature domain context as condition. This enables us to leverage the high dimension context to carry rich information to both the encoder and the decoder, which helps reconstruct the high-frequency contents for higher video quality. Our framework is also extensible, in which the condition can be flexibly designed. Experiments show that our method can significantly outperform the previous state-of-the-art (SOTA) deep video compression methods. When compared with x265 using veryslow preset, we can achieve 26.0% bitrate saving for 1080P standard test videos.

arXiv:2109.14837

End-to-End Image Compression with Probabilistic Decoding

Lossy image compression is a many-to-one process, thus one bitstream corresponds to multiple possible original images, especially at low bit rates. However, this nature was seldom considered in previous studies on image compression, which usually chose one possible image as reconstruction, e.g. the one with the maximal a posteriori probability. We propose a learned image compression framework to natively support probabilistic decoding. The compressed bitstream is decoded into a series of parameters that instantiate a pre-chosen distribution; then the distribution is used by the decoder to sample and reconstruct images. The decoder may adopt different sampling strategies and produce diverse reconstructions, among which some have higher signal fidelity and some others have better visual quality. The proposed framework is dependent on a revertible neural network-based transform to convert pixels into coefficients that obey the pre-chosen distribution as much as possible. Our code and models will be made publicly available.

arXiv:2109.14818공식 코드 · Papers with Code

Spark in the Dark: Evaluating Encoder-Decoder Pairs for COVID-19 CT's Semantic Segmentation

With the COVID-19 global pandemic, computerassisted diagnoses of medical images have gained a lot of attention, and robust methods of Semantic Segmentation of Computed Tomography (CT) turned highly desirable. Semantic Segmentation of CT is one of many research fields of automatic detection of Covid-19 and was widely explored since the Covid19 outbreak. In the robotic field, Semantic Segmentation of organs and CTs are widely used in robots developed for surgery tasks. As new methods and new datasets are proposed quickly, it becomes apparent the necessity of providing an extensive evaluation of those methods. To provide a standardized comparison of different architectures across multiple recently proposed datasets, we propose in this paper an extensive benchmark of multiple encoders and decoders with a total of 120 architectures evaluated in five datasets, with each dataset being validated through a five-fold cross-validation strategy, totaling 3.000 experiments. To the best of our knowledge, this is the largest evaluation in number of encoders, decoders, and datasets proposed in the field of Covid-19 CT segmentation.

arXiv:2109.14657

Understanding Egocentric Hand-Object Interactions from Hand Pose Estimation

In this paper, we address the problem of estimating the hand pose from the egocentric view when the hand is interacting with objects. Specifically, we propose a method to label a dataset Ego-Siam which contains the egocentric images pair-wisely. We also use the collected pairwise data to train our encoder-decoder style network which has been proven efficient in. This could bring extra training efficiency and testing accuracy. Our network is lightweight and can be performed with over 30 FPS with an outdated GPU. We demonstrate that our method outperforms Mueller et al. which is the state of the art work dealing with egocentric hand-object interaction problems on the GANerated dataset. To show the ability to preserve the semantic information of our method, we also report the performance of grasp type classification on GUN-71 dataset and outperforms the benchmark by only using the predicted 3-d hand pose.

arXiv:2109.15149공식 코드 · Papers with Code

Deep Embedded K-Means Clustering

Recently, deep clustering methods have gained momentum because of the high representational power of deep neural networks (DNNs) such as autoencoder. The key idea is that representation learning and clustering can reinforce each other: Good representations lead to good clustering while good clustering provides good supervisory signals to representation learning. Critical questions include: 1) How to optimize representation learning and clustering? 2) Should the reconstruction loss of autoencoder be considered always? In this paper, we propose DEKM (for Deep Embedded K-Means) to answer these two questions. Since the embedding space generated by autoencoder may have no obvious cluster structures, we propose to further transform the embedding space to a new space that reveals the cluster-structure information. This is achieved by an orthonormal transformation matrix, which contains the eigenvectors of the within-class scatter matrix of K-means. The eigenvalues indicate the importance of the eigenvectors' contributions to the cluster-structure information in the new space. Our goal is to increase the cluster-structure information. To this end, we discard the decoder and propose a greedy method to optimize the representation. Representation learning and clustering are alternately optimized by DEKM. Experimental results on the real-world datasets demonstrate that DEKM achieves state-of-the-art performance.

arXiv:2109.14530공식 코드 · Papers with Code

Deep Spatio-Temporal Wind Power Forecasting

Wind power forecasting has drawn increasing attention among researchers as the consumption of renewable energy grows. In this paper, we develop a deep learning approach based on encoder-decoder structure. Our model forecasts wind power generated by a wind turbine using its spatial location relative to other turbines and historical wind speed data. In this way, we effectively integrate spatial dependency and temporal trends to make turbine-specific predictions. The advantages of our method over existing work can be summarized as 1) it directly predicts wind power based on historical wind speed, without the need for prediction of wind speed first, and then using a transformation; 2) it can effectively capture long-term dependency 3) our model is more scalable and efficient compared with other deep learning based methods. We demonstrate the efficacy of our model on the benchmark real-world datasets.

arXiv:2109.14420공식 코드 · Papers with Code

FastCorrect 2: Fast Error Correction on Multiple Candidates for Automatic Speech Recognition

Error correction is widely used in automatic speech recognition (ASR) to post-process the generated sentence, and can further reduce the word error rate (WER). Although multiple candidates are generated by an ASR system through beam search, current error correction approaches can only correct one sentence at a time, failing to leverage the voting effect from multiple candidates to better detect and correct error tokens. In this work, we propose FastCorrect 2, an error correction model that takes multiple ASR candidates as input for better correction accuracy. FastCorrect 2 adopts non-autoregressive generation for fast inference, which consists of an encoder that processes multiple source sentences and a decoder that generates the target sentence in parallel from the adjusted source sentence, where the adjustment is based on the predicted duration of each source token. However, there are some issues when handling multiple source sentences. First, it is non-trivial to leverage the voting effect from multiple source sentences since they usually vary in length. Thus, we propose a novel alignment algorithm to maximize the degree of token alignment among multiple sentences in terms of token and pronunciation similarity. Second, the decoder can only take one adjusted source sentence as input, while there are multiple source sentences. Thus, we develop a candidate predictor to detect the most suitable candidate for the decoder. Experiments on our inhouse dataset and AISHELL-1 show that FastCorrect 2 can further reduce the WER over the previous correction model with single candidate by 3.2% and 2.6%, demonstrating the effectiveness of leveraging multiple candidates in ASR error correction. FastCorrect 2 achieves better performance than the cascaded re-scoring and correction pipeline and can serve as a unified post-processing module for ASR.

arXiv:2109.14249

Multipath CNN with alpha matte inference for knee tissue segmentation from MRI

Precise segmentation of knee tissues from magnetic resonance imaging (MRI) is critical in quantitative imaging and diagnosis. Convolutional neural networks (CNNs), which are state of the art, have limitations owing to the lack of image-specific adaptation, such as low tissue contrasts and structural inhomogeneities, thereby leading to incomplete segmentation results. This paper presents a deep learning based automatic segmentation framework for knee tissue segmentation. A novel multipath CNN-based method is proposed, which consists of an encoder decoder-based segmentation network in combination with a low rank tensor-reconstructed segmentation network. Low rank reconstruction in MRI tensor sub-blocks is introduced to exploit the structural and morphological variations in knee tissues. To further improve the segmentation from CNNs, trimap generation, which effectively utilizes superimposed regions, is proposed for defining high, medium and low confidence regions from the multipath CNNs. The secondary path with low rank reconstructed input mitigates the conditions in which the primary segmentation network can potentially fail and overlook the boundary regions. The outcome of the segmentation is solved as an alpha matting problem by blending the trimap with the source input. Experiments on Osteoarthritis Initiative (OAI) datasets and a self prepared scan validate the effectiveness of the proposed method. We specifically demonstrate the application of the proposed method in a cartilage segmentation based thickness map for diagnosis purposes.

arXiv:2109.13858

Domain Generalization for Vision-based Driving Trajectory Generation

One of the challenges in vision-based driving trajectory generation is dealing with out-of-distribution scenarios. In this paper, we propose a domain generalization method for vision-based driving trajectory generation for autonomous vehicles in urban environments, which can be seen as a solution to extend the Invariant Risk Minimization (IRM) method in complex problems. We leverage an adversarial learning approach to train a trajectory generator as the decoder. Based on the pre-trained decoder, we infer the latent variables corresponding to the trajectories, and pre-train the encoder by regressing the inferred latent variable. Finally, we fix the decoder but fine-tune the encoder with the final trajectory loss. We compare our proposed method with the state-of-the-art trajectory generation method and some recent domain generalization methods on both datasets and simulation, demonstrating that our method has better generalization ability.