PyTorchKR Papers

최신 논문

총 9,339편
태그·기간 등 상세 필터

카테고리 탐색

로그인 후 이 태그 구독

arXiv:2305.16699공식 코드 · Papers with Code

Automatic Tuning of Loss Trade-offs without Hyper-parameter Search in End-to-End Zero-Shot Speech Synthesis

Recently, zero-shot TTS and VC methods have gained attention due to their practicality of being able to generate voices even unseen during training. Among these methods, zero-shot modifications of the VITS model have shown superior performance, while having useful properties inherited from VITS. However, the performance of VITS and VITS-based zero-shot models vary dramatically depending on how the losses are balanced. This can be problematic, as it requires a burdensome procedure of tuning loss balance hyper-parameters to find the optimal balance. In this work, we propose a novel framework that finds this optimum without search, by inducing the decoder of VITS-based models to its full reconstruction ability. With our framework, we show superior performance compared to baselines in zero-shot TTS and VC, achieving state-of-the-art performance. Furthermore, we show the robustness of our framework in various settings. We provide an explanation for the results in the discussion.

arXiv:2305.16620

Pedestrian Trajectory Forecasting Using Deep Ensembles Under Sensing Uncertainty

One of the fundamental challenges in the prediction of dynamic agents is robustness. Usually, most predictions are deterministic estimates of future states which are over-confident and prone to error. Recently, few works have addressed capturing uncertainty during forecasting of future states. However, these probabilistic estimation methods fail to account for the upstream noise in perception data during tracking. Sensors always have noise and state estimation becomes even more difficult under adverse weather conditions and occlusion. Traditionally, Bayes filters have been used to fuse information from noisy sensors to update states with associated belief. But, they fail to address non-linearities and long-term predictions. Therefore, we propose an end-to-end estimator that can take noisy sensor measurements and make robust future state predictions with uncertainty bounds while simultaneously taking into consideration the upstream perceptual uncertainty. For the current research, we consider an encoder-decoder based deep ensemble network for capturing both perception and predictive uncertainty simultaneously. We compared the current model to other approximate Bayesian inference methods. Overall, deep ensembles provided more robust predictions and the consideration of upstream uncertainty further increased the estimation accuracy for the model.

arXiv:2305.16548공식 코드 · Papers with Code

Annotating and Detecting Fine-grained Factual Errors for Dialogue Summarization

A series of datasets and models have been proposed for summaries generated for well-formatted documents such as news articles. Dialogue summaries, however, have been under explored. In this paper, we present the first dataset with fine-grained factual error annotations named DIASUMFACT. We define fine-grained factual error detection as a sentence-level multi-label classification problem, and we evaluate two state-of-the-art (SOTA) models on our dataset. Both models yield sub-optimal results, with a macro-averaged F1 score of around 0.25 over 6 error classes. We further propose an unsupervised model ENDERANKER via candidate ranking using pretrained encoder-decoder models. Our model performs on par with the SOTA models while requiring fewer resources. These observations confirm the challenges in detecting factual errors from dialogue summaries, which call for further studies, for which our dataset and results offer a solid foundation.

arXiv:2305.16380

Scan and Snap: Understanding Training Dynamics and Token Composition in 1-layer Transformer

Transformer architecture has shown impressive performance in multiple research domains and has become the backbone of many neural network models. However, there is limited understanding on how it works. In particular, with a simple predictive loss, how the representation emerges from the gradient \emph{training dynamics} remains a mystery. In this paper, for 1-layer transformer with one self-attention layer plus one decoder layer, we analyze its SGD training dynamics for the task of next token prediction in a mathematically rigorous manner. We open the black box of the dynamic process of how the self-attention layer combines input tokens, and reveal the nature of underlying inductive bias. More specifically, with the assumption (a) no positional encoding, (b) long input sequence, and (c) the decoder layer learns faster than the self-attention layer, we prove that self-attention acts as a \emph{discriminative scanning algorithm}: starting from uniform attention, it gradually attends more to distinct key tokens for a specific next token to be predicted, and pays less attention to common key tokens that occur across different next tokens. Among distinct tokens, it progressively drops attention weights, following the order of low to high co-occurrence between the key and the query token in the training set. Interestingly, this procedure does not lead to winner-takes-all, but decelerates due to a \emph{phase transition} that is controllable by the learning rates of the two layers, leaving (almost) fixed token combination. We verify this \textbf{\emph{scan and snap}} dynamics on synthetic and real-world data (WikiText).

arXiv:2305.15958

Improving Scheduled Sampling for Neural Transducer-based ASR

The recurrent neural network-transducer (RNNT) is a promising approach for automatic speech recognition (ASR) with the introduction of a prediction network that autoregressively considers linguistic aspects. To train the autoregressive part, the ground-truth tokens are used as substitutions for the previous output token, which leads to insufficient robustness to incorrect past tokens; a recognition error in the decoding leads to further errors. Scheduled sampling (SS) is a technique to train autoregressive model robustly to past errors by randomly replacing some ground-truth tokens with actual outputs generated from a model. SS mitigates the gaps between training and decoding steps, known as exposure bias, and it is often used for attentional encoder-decoder training. However SS has not been fully examined for RNNT because of the difficulty in applying SS to RNNT due to the complicated RNNT output form. In this paper we propose SS approaches suited for RNNT. Our SS approaches sample the tokens generated from the distiribution of RNNT itself, i.e. internal language model or RNNT outputs. Experiments in three datasets confirm that RNNT trained with our SS approach achieves the best ASR performance. In particular, on a Japanese ASR task, our best system outperforms the previous state-of-the-art alternative.

arXiv:2305.16155공식 코드 · Papers with Code

Revisiting Non-Autoregressive Translation at Scale

In real-world systems, scaling has been critical for improving the translation quality in autoregressive translation (AT), which however has not been well studied for non-autoregressive translation (NAT). In this work, we bridge the gap by systematically studying the impact of scaling on NAT behaviors. Extensive experiments on six WMT benchmarks over two advanced NAT models show that scaling can alleviate the commonly-cited weaknesses of NAT models, resulting in better translation performance. To reduce the side-effect of scaling on decoding speed, we empirically investigate the impact of NAT encoder and decoder on the translation performance. Experimental results on the large-scale WMT20 En-De show that the asymmetric architecture (e.g. bigger encoder and smaller decoder) can achieve comparable performance with the scaling model, while maintaining the superiority of decoding speed with standard NAT models. To this end, we establish a new benchmark by validating scaled NAT models on the scaled dataset, which can be regarded as a strong baseline for future works. We release code and system outputs at https://github.com/DeepLearnXMU/Scaling4NAT.

arXiv:2305.16107

VioLA: Unified Codec Language Models for Speech Recognition, Synthesis, and Translation

Recent research shows a big convergence in model architecture, training objectives, and inference methods across various tasks for different modalities. In this paper, we propose VioLA, a single auto-regressive Transformer decoder-only network that unifies various cross-modal tasks involving speech and text, such as speech-to-text, text-to-text, text-to-speech, and speech-to-speech tasks, as a conditional codec language model task via multi-task learning framework. To accomplish this, we first convert all the speech utterances to discrete tokens (similar to the textual data) using an offline neural codec encoder. In such a way, all these tasks are converted to token-based sequence conversion problems, which can be naturally handled with one conditional language model. We further integrate task IDs (TID) and language IDs (LID) into the proposed model to enhance the modeling capability of handling different languages and tasks. Experimental results demonstrate that the proposed VioLA model can support both single-modal and cross-modal tasks well, and the decoder-only model achieves a comparable and even better performance than the strong baselines.

arXiv:2305.16319

Image as First-Order Norm+Linear Autoregression: Unveiling Mathematical Invariance

This paper introduces a novel mathematical property applicable to diverse images, referred to as FINOLA (First-Order Norm+Linear Autoregressive). FINOLA represents each image in the latent space as a first-order autoregressive process, in which each regression step simply applies a shared linear model on the normalized value of its immediate neighbor. This intriguing property reveals a mathematical invariance that transcends individual images. Expanding from image grids to continuous coordinates, we unveil the presence of two underlying partial differential equations. We validate the FINOLA property from two distinct angles: image reconstruction and self-supervised learning. Firstly, we demonstrate the ability of FINOLA to auto-regress up to a 256x256 feature map (the same resolution to the image) from a single vector placed at the center, successfully reconstructing the original image by only using three 3x3 convolution layers as decoder. Secondly, we leverage FINOLA for self-supervised learning by employing a simple masked prediction approach. Encoding a single unmasked quadrant block, we autoregressively predict the surrounding masked region. Remarkably, this pre-trained representation proves highly effective in image classification and object detection tasks, even when integrated into lightweight networks, all without the need for extensive fine-tuning. The code will be made publicly available.

arXiv:2305.16233

Interactive Segment Anything NeRF with Feature Imitation

This paper investigates the potential of enhancing Neural Radiance Fields (NeRF) with semantics to expand their applications. Although NeRF has been proven useful in real-world applications like VR and digital creation, the lack of semantics hinders interaction with objects in complex scenes. We propose to imitate the backbone feature of off-the-shelf perception models to achieve zero-shot semantic segmentation with NeRF. Our framework reformulates the segmentation process by directly rendering semantic features and only applying the decoder from perception models. This eliminates the need for expensive backbones and benefits 3D consistency. Furthermore, we can project the learned semantics onto extracted mesh surfaces for real-time interaction. With the state-of-the-art Segment Anything Model (SAM), our framework accelerates segmentation by 16 times with comparable mask quality. The experimental results demonstrate the efficacy and computational advantages of our approach. Project page: \url{https://me.kiui.moe/san/}.

arXiv:2305.16172공식 코드 · Papers with Code

Masked and Permuted Implicit Context Learning for Scene Text Recognition

Scene Text Recognition (STR) is difficult because of the variations in text styles, shapes, and backgrounds. Though the integration of linguistic information enhances models' performance, existing methods based on either permuted language modeling (PLM) or masked language modeling (MLM) have their pitfalls. PLM's autoregressive decoding lacks foresight into subsequent characters, while MLM overlooks inter-character dependencies. Addressing these problems, we propose a masked and permuted implicit context learning network for STR, which unifies PLM and MLM within a single decoder, inheriting the advantages of both approaches. We utilize the training procedure of PLM, and to integrate MLM, we incorporate word length information into the decoding process and replace the undetermined characters with mask tokens. Besides, perturbation training is employed to train a more robust model against potential length prediction errors. Our empirical evaluations demonstrate the performance of our model. It not only achieves superior performance on the common benchmarks but also achieves a substantial improvement of $9.1\%$ on the more challenging Union14M-Benchmark.

arXiv:2305.16195

Abstractive Summary Generation for the Urdu Language

Abstractive summary generation is a challenging task that requires the model to comprehend the source text and generate a concise and coherent summary that captures the essential information. In this paper, we explore the use of an encoder/decoder approach for abstractive summary generation in the Urdu language. We employ a transformer-based model that utilizes self-attention mechanisms to encode the input text and generate a summary. Our experiments show that our model can produce summaries that are grammatically correct and semantically meaningful. We evaluate our model on a publicly available dataset and achieve state-of-the-art results in terms of Rouge scores. We also conduct a qualitative analysis of our model's output to assess its effectiveness and limitations. Our findings suggest that the encoder/decoder approach is a promising method for abstractive summary generation in Urdu and can be extended to other languages with suitable modifications.

arXiv:2305.16295

HAAV: Hierarchical Aggregation of Augmented Views for Image Captioning

A great deal of progress has been made in image captioning, driven by research into how to encode the image using pre-trained models. This includes visual encodings (e.g. image grid features or detected objects) and more recently textual encodings (e.g. image tags or text descriptions of image regions). As more advanced encodings are available and incorporated, it is natural to ask: how to efficiently and effectively leverage the heterogeneous set of encodings? In this paper, we propose to regard the encodings as augmented views of the input image. The image captioning model encodes each view independently with a shared encoder efficiently, and a contrastive loss is incorporated across the encoded views in a novel way to improve their representation quality and the model's data efficiency. Our proposed hierarchical decoder then adaptively weighs the encoded views according to their effectiveness for caption generation by first aggregating within each view at the token level, and then across views at the view level. We demonstrate significant performance improvements of +5.6% CIDEr on MS-COCO and +12.9% CIDEr on Flickr30k compared to state of the arts, and conduct rigorous analyses to demonstrate the importance of each part of our design.

arXiv:2305.15404공식 코드 · Papers with Code

RoMa: Robust Dense Feature Matching

Feature matching is an important computer vision task that involves estimating correspondences between two images of a 3D scene, and dense methods estimate all such correspondences. The aim is to learn a robust model, i.e., a model able to match under challenging real-world changes. In this work, we propose such a model, leveraging frozen pretrained features from the foundation model DINOv2. Although these features are significantly more robust than local features trained from scratch, they are inherently coarse. We therefore combine them with specialized ConvNet fine features, creating a precisely localizable feature pyramid. To further improve robustness, we propose a tailored transformer match decoder that predicts anchor probabilities, which enables it to express multimodality. Finally, we propose an improved loss formulation through regression-by-classification with subsequent robust regression. We conduct a comprehensive set of experiments that show that our method, RoMa, achieves significant gains, setting a new state-of-the-art. In particular, we achieve a 36% improvement on the extremely challenging WxBS benchmark. Code is provided at https://github.com/Parskatt/RoMa

arXiv:2305.15302

Multi-Modal Mutual Attention and Iterative Interaction for Referring Image Segmentation

We address the problem of referring image segmentation that aims to generate a mask for the object specified by a natural language expression. Many recent works utilize Transformer to extract features for the target object by aggregating the attended visual regions. However, the generic attention mechanism in Transformer only uses the language input for attention weight calculation, which does not explicitly fuse language features in its output. Thus, its output feature is dominated by vision information, which limits the model to comprehensively understand the multi-modal information, and brings uncertainty for the subsequent mask decoder to extract the output mask. To address this issue, we propose Multi-Modal Mutual Attention ($\mathrm{M^3Att}$) and Multi-Modal Mutual Decoder ($\mathrm{M^3Dec}$) that better fuse information from the two input modalities. Based on {$\mathrm{M^3Dec}$}, we further propose Iterative Multi-modal Interaction ($\mathrm{IMI}$) to allow continuous and in-depth interactions between language and vision features. Furthermore, we introduce Language Feature Reconstruction ($\mathrm{LFR}$) to prevent the language information from being lost or distorted in the extracted feature. Extensive experiments show that our proposed approach significantly improves the baseline and outperforms state-of-the-art referring image segmentation methods on RefCOCO series datasets consistently.

arXiv:2305.14955공식 코드 · Papers with Code

DC-Net: Divide-and-Conquer for Salient Object Detection

In this paper, we introduce Divide-and-Conquer into the salient object detection (SOD) task to enable the model to learn prior knowledge that is for predicting the saliency map. We design a novel network, Divide-and-Conquer Network (DC-Net) which uses two encoders to solve different subtasks that are conducive to predicting the final saliency map, here is to predict the edge maps with width 4 and location maps of salient objects and then aggregate the feature maps with different semantic information into the decoder to predict the final saliency map. The decoder of DC-Net consists of our newly designed two-level Residual nested-ASPP (ResASPP$^{2}$) modules, which have the ability to capture a large number of different scale features with a small number of convolution operations and have the advantages of maintaining high resolution all the time and being able to obtain a large and compact effective receptive field (ERF). Based on the advantage of Divide-and-Conquer's parallel computing, we use Parallel Acceleration to speed up DC-Net, allowing it to achieve competitive performance on six LR-SOD and five HR-SOD datasets under high efficiency (60 FPS and 55 FPS). Codes and results are available: https://github.com/PiggyJerry/DC-Net.

arXiv:2305.14951공식 코드 · Papers with Code

DSFFNet: Dual-Side Feature Fusion Network for 3D Pose Transfer

To solve the problem of pose distortion in the forward propagation of pose features in existing methods, this pa-per proposes a Dual-Side Feature Fusion Network for pose transfer (DSFFNet). Firstly, a fixed-length pose code is extracted from the source mesh by a pose encoder and combined with the target vertices to form a mixed feature; Then, a Feature Fusion Adaptive Instance Normalization module (FFAdaIN) is designed, which can process both pose and identity features simultaneously, so that the pose features can be compensated in layer-by-layer for-ward propagation, thus solving the pose distortion problem; Finally, using the mesh decoder composed of this module, the pose are gradually transferred to the target mesh. Experimental results on SMPL, SMAL, FAUST and MultiGarment datasets show that DSFFNet successfully solves the pose distortion problem while maintaining a smaller network structure with stronger pose transfer capability and faster convergence speed, and can adapt to meshes with different numbers of vertices. Code is available at https://github.com/YikiDragon/DSFFNet

arXiv:2305.14684공식 코드 · Papers with Code

Collaborative Auto-encoding for Blind Image Quality Assessment

Blind image quality assessment (BIQA) is a challenging problem with important real-world applications. Recent efforts attempting to exploit powerful representations by deep neural networks (DNN) are hindered by the lack of subjectively annotated data. This paper presents a novel BIQA method which overcomes this fundamental obstacle. Specifically, we design a pair of collaborative autoencoders (COAE) consisting of a content autoencoder (CAE) and a distortion autoencoder (DAE) that work together to extract content and distortion representations, which are shown to be highly descriptive of image quality. While the CAE follows a standard codec procedure, we introduce the CAE-encoded feature as an extra input to the DAE's decoder for reconstructing distorted images, thus effectively forcing DAE's encoder to extract distortion representations. The self-supervised learning framework allows the COAE including two feature extractors to be trained by almost unlimited amount of data, thus leaving limited samples with annotations to finetune a BIQA model. We will show that the proposed BIQA method achieves state-of-the-art performance and has superior generalization capability over other learning based models. The codes are available at: https://github.com/Macro-Zhou/NRIQA-VISOR/.

arXiv:2305.15025공식 코드 · Papers with Code

Dior-CVAE: Pre-trained Language Models and Diffusion Priors for Variational Dialog Generation

Current variational dialog models have employed pre-trained language models (PLMs) to parameterize the likelihood and posterior distributions. However, the Gaussian assumption made on the prior distribution is incompatible with these distributions, thus restricting the diversity of generated responses. These models also suffer from posterior collapse, i.e., the decoder tends to ignore latent variables and directly access information captured in the encoder through the cross-attention mechanism. In this work, we propose Dior-CVAE, a hierarchical conditional variational autoencoder (CVAE) with diffusion priors to address these challenges. We employ a diffusion model to increase the complexity of the prior distribution and its compatibility with the distributions produced by a PLM. Also, we propose memory dropout to the cross-attention mechanism, which actively encourages the use of latent variables for response generation. Overall, experiments across two commonly used open-domain dialog datasets show that our method can generate more diverse responses without large-scale dialog pre-training. Code is available at https://github.com/UKPLab/dior-cvae.

arXiv:2305.14949

Cross-lingual Data Augmentation for Document-grounded Dialog Systems in Low Resource Languages

This paper proposes a framework to address the issue of data scarcity in Document-Grounded Dialogue Systems(DGDS). Our model leverages high-resource languages to enhance the capability of dialogue generation in low-resource languages. Specifically, We present a novel pipeline CLEM (Cross-Lingual Enhanced Model) including adversarial training retrieval (Retriever and Re-ranker), and Fid (fusion-in-decoder) generator. To further leverage high-resource language, we also propose an innovative architecture to conduct alignment across different languages with translated training. Extensive experiment results demonstrate the effectiveness of our model and we achieved 4th place in the DialDoc 2023 Competition. Therefore, CLEM can serve as a solution to resource scarcity in DGDS and provide useful guidance for multi-lingual alignment tasks.

arXiv:2305.14761공식 코드 · Papers with Code

UniChart: A Universal Vision-language Pretrained Model for Chart Comprehension and Reasoning

Charts are very popular for analyzing data, visualizing key insights and answering complex reasoning questions about data. To facilitate chart-based data analysis using natural language, several downstream tasks have been introduced recently such as chart question answering and chart summarization. However, most of the methods that solve these tasks use pretraining on language or vision-language tasks that do not attempt to explicitly model the structure of the charts (e.g., how data is visually encoded and how chart elements are related to each other). To address this, we first build a large corpus of charts covering a wide variety of topics and visual styles. We then present UniChart, a pretrained model for chart comprehension and reasoning. UniChart encodes the relevant text, data, and visual elements of charts and then uses a chart-grounded text decoder to generate the expected output in natural language. We propose several chart-specific pretraining tasks that include: (i) low-level tasks to extract the visual elements (e.g., bars, lines) and data from charts, and (ii) high-level tasks to acquire chart understanding and reasoning skills. We find that pretraining the model on a large corpus with chart-specific low- and high-level tasks followed by finetuning on three down-streaming tasks results in state-of-the-art performance on three downstream tasks.