PyTorchKR Papers

최신 논문

총 5,358편
태그·기간 등 상세 필터

카테고리 탐색

로그인 후 이 태그 구독

arXiv:2103.15432

Towards High Fidelity Monocular Face Reconstruction with Rich Reflectance using Self-supervised Learning and Ray Tracing

Robust face reconstruction from monocular image in general lighting conditions is challenging. Methods combining deep neural network encoders with differentiable rendering have opened up the path for very fast monocular reconstruction of geometry, lighting and reflectance. They can also be trained in self-supervised manner for increased robustness and better generalization. However, their differentiable rasterization based image formation models, as well as underlying scene parameterization, limit them to Lambertian face reflectance and to poor shape details. More recently, ray tracing was introduced for monocular face reconstruction within a classic optimization-based framework and enables state-of-the art results. However optimization-based approaches are inherently slow and lack robustness. In this paper, we build our work on the aforementioned approaches and propose a new method that greatly improves reconstruction quality and robustness in general scenes. We achieve this by combining a CNN encoder with a differentiable ray tracer, which enables us to base the reconstruction on much more advanced personalized diffuse and specular albedos, a more sophisticated illumination model and a plausible representation of self-shadows. This enables to take a big leap forward in reconstruction quality of shape, appearance and lighting even in scenes with difficult illumination. With consistent face attributes reconstruction, our method leads to practical applications such as relighting and self-shadows removal. Compared to state-of-the-art methods, our results show improved accuracy and validity of the approach.

arXiv:2103.15578

Classification of Seeds using Domain Randomization on Self-Supervised Learning Frameworks

The first step toward Seed Phenotyping i.e. the comprehensive assessment of complex seed traits such as growth, development, tolerance, resistance, ecology, yield, and the measurement of pa-rameters that form more complex traits is the identification of seed type. Generally, a plant re-searcher inspects the visual attributes of a seed such as size, shape, area, color and texture to identify the seed type, a process that is tedious and labor-intensive. Advances in the areas of computer vision and deep learning have led to the development of convolutional neural networks (CNN) that aid in classification using images. While they classify efficiently, a key bottleneck is the need for an extensive amount of labelled data to train the CNN before it can be put to the task of classification. The work leverages the concepts of Contrastive Learning and Domain Randomi-zation in order to achieve the same. Briefly, domain randomization is the technique of applying models trained on images containing simulated objects to real-world objects. The use of synthetic images generated from a representational sample crop of real-world images alleviates the need for a large volume of test subjects. As part of the work, synthetic image datasets of five different types of seed images namely, canola, rough rice, sorghum, soy and wheat are applied to three different self-supervised learning frameworks namely, SimCLR, Momentum Contrast (MoCo) and Build Your Own Latent (BYOL) where ResNet-50 is used as the backbone in each of the networks. When the self-supervised models are fine-tuned with only 5% of the labels from the synthetic dataset, results show that MoCo, the model that yields the best performance of the self-supervised learning frameworks in question, achieves an accuracy of 77% on the test dataset which is only ~13% less than the accuracy of 90% achieved by ResNet-50 trained on 100% of the labels.

arXiv:2103.14542

Unsupervised Document Embedding via Contrastive Augmentation

We present a contrasting learning approach with data augmentation techniques to learn document representations in an unsupervised manner. Inspired by recent contrastive self-supervised learning algorithms used for image and NLP pretraining, we hypothesize that high-quality document embedding should be invariant to diverse paraphrases that preserve the semantics of the original document. With different backbones and contrastive learning frameworks, our study reveals the enormous benefits of contrastive augmentation for document representation learning with two additional insights: 1) including data augmentation in a contrastive way can substantially improve the embedding quality in unsupervised document representation learning, and 2) in general, stochastic augmentations generated by simple word-level manipulation work much better than sentence-level and document-level ones. We plug our method into a classifier and compare it with a broad range of baseline methods on six benchmark datasets. Our method can decrease the classification error rate by up to 6.4% over the SOTA approaches on the document classification task, matching or even surpassing fully-supervised methods.

arXiv:2103.14417공식 코드 · Papers with Code

Self-Supervised Learning in Multi-Task Graphs through Iterative Consensus Shift

The human ability to synchronize the feedback from all their senses inspired recent works in multi-task and multi-modal learning. While these works rely on expensive supervision, our multi-task graph requires only pseudo-labels from expert models. Every graph node represents a task, and each edge learns between tasks transformations. Once initialized, the graph learns self-supervised, based on a novel consensus shift algorithm that intelligently exploits the agreement between graph pathways to generate new pseudo-labels for the next learning cycle. We demonstrate significant improvement from one unsupervised learning iteration to the next, outperforming related recent methods in extensive multi-task learning experiments on two challenging datasets. Our code is available at https://github.com/bit-ml/cshift.

arXiv:2103.14152

Residual Energy-Based Models for End-to-End Speech Recognition

End-to-end models with auto-regressive decoders have shown impressive results for automatic speech recognition (ASR). These models formulate the sequence-level probability as a product of the conditional probabilities of all individual tokens given their histories. However, the performance of locally normalised models can be sub-optimal because of factors such as exposure bias. Consequently, the model distribution differs from the underlying data distribution. In this paper, the residual energy-based model (R-EBM) is proposed to complement the auto-regressive ASR model to close the gap between the two distributions. Meanwhile, R-EBMs can also be regarded as utterance-level confidence estimators, which may benefit many downstream tasks. Experiments on a 100hr LibriSpeech dataset show that R-EBMs can reduce the word error rates (WERs) by 8.2%/6.7% while improving areas under precision-recall curves of confidence scores by 12.6%/28.4% on test-clean/test-other sets. Furthermore, on a state-of-the-art model using self-supervised learning (wav2vec 2.0), R-EBMs still significantly improves both the WER and confidence estimation performance.

arXiv:2103.13559공식 코드 · Papers with Code

Rethinking Self-Supervised Learning: Small is Beautiful

Self-supervised learning (SSL), in particular contrastive learning, has made great progress in recent years. However, a common theme in these methods is that they inherit the learning paradigm from the supervised deep learning scenario. Current SSL methods are often pretrained for many epochs on large-scale datasets using high resolution images, which brings heavy computational cost and lacks flexibility. In this paper, we demonstrate that the learning paradigm for SSL should be different from supervised learning and the information encoded by the contrastive loss is expected to be much less than that encoded in the labels in supervised learning via the cross entropy loss. Hence, we propose scaled-down self-supervised learning (S3L), which include 3 parts: small resolution, small architecture and small data. On a diverse set of datasets, SSL methods and backbone architectures, S3L achieves higher accuracy consistently with much less training cost when compared to previous SSL learning paradigm. Furthermore, we show that even without a large pretraining dataset, S3L can achieve impressive results on small data alone. Our code has been made publically available at https://github.com/CupidJay/Scaled-down-self-supervised-learning.

arXiv:2103.13716

Vectorization and Rasterization: Self-Supervised Learning for Sketch and Handwriting

Self-supervised learning has gained prominence due to its efficacy at learning powerful representations from unlabelled data that achieve excellent performance on many challenging downstream tasks. However supervision-free pre-text tasks are challenging to design and usually modality specific. Although there is a rich literature of self-supervised methods for either spatial (such as images) or temporal data (sound or text) modalities, a common pre-text task that benefits both modalities is largely missing. In this paper, we are interested in defining a self-supervised pre-text task for sketches and handwriting data. This data is uniquely characterised by its existence in dual modalities of rasterized images and vector coordinate sequences. We address and exploit this dual representation by proposing two novel cross-modal translation pre-text tasks for self-supervised feature learning: Vectorization and Rasterization. Vectorization learns to map image space to vector coordinates and rasterization maps vector coordinates to image space. We show that the our learned encoder modules benefit both raster-based and vector-based downstream approaches to analysing hand-drawn data. Empirical evidence shows that our novel pre-text tasks surpass existing single and multi-modal self-supervision methods.

arXiv:2103.13125

Learning Graph Representation by Aggregating Subgraphs via Mutual Information Maximization

In this paper, we introduce a self-supervised learning method to enhance the graph-level representations with the help of a set of subgraphs. For this purpose, we propose a universal framework to generate subgraphs in an auto-regressive way and then using these subgraphs to guide the learning of graph representation by Graph Neural Networks. Under this framework, we can get a comprehensive understanding of the graph structure in a learnable way. And to fully capture enough information of original graphs, we design three information aggregators: \textbf{attribute-conv}, \textbf{layer-conv} and \textbf{subgraph-conv} to gather information from different aspects. And to achieve efficient and effective contrastive learning, a Head-Tail contrastive construction is proposed to provide abundant negative samples. Under all proposed components which can be generalized to any Graph Neural Networks, in the unsupervised case, we achieve new state-of-the-art results in several benchmarks. We also evaluate our model on semi-supervised learning tasks and make a fair comparison to state-of-the-art semi-supervised methods.

arXiv:2103.13047

Knowledge-aware Contrastive Molecular Graph Learning

Leveraging domain knowledge including fingerprints and functional groups in molecular representation learning is crucial for chemical property prediction and drug discovery. When modeling the relation between graph structure and molecular properties implicitly, existing works can hardly capture structural or property changes and complex structure, with much smaller atom vocabulary and highly frequent atoms. In this paper, we propose the Contrastive Knowledge-aware GNN (CKGNN) for self-supervised molecular representation learning to fuse domain knowledge into molecular graph representation. We explicitly encode domain knowledge via knowledge-aware molecular encoder under the contrastive learning framework, ensuring that the generated molecular embeddings equipped with chemical domain knowledge to distinguish molecules with similar chemical formula but dissimilar functions. Extensive experiments on 8 public datasets demonstrate the effectiveness of our model with a 6\% absolute improvement on average against strong competitors. Ablation study and further investigation also verify the best of both worlds: incorporation of chemical domain knowledge into self-supervised learning.

arXiv:2103.13061공식 코드 · Papers with Code

Revamping Cross-Modal Recipe Retrieval with Hierarchical Transformers and Self-supervised Learning

Cross-modal recipe retrieval has recently gained substantial attention due to the importance of food in people's lives, as well as the availability of vast amounts of digital cooking recipes and food images to train machine learning models. In this work, we revisit existing approaches for cross-modal recipe retrieval and propose a simplified end-to-end model based on well established and high performing encoders for text and images. We introduce a hierarchical recipe Transformer which attentively encodes individual recipe components (titles, ingredients and instructions). Further, we propose a self-supervised loss function computed on top of pairs of individual recipe components, which is able to leverage semantic relationships within recipes, and enables training using both image-recipe and recipe-only samples. We conduct a thorough analysis and ablation studies to validate our design choices. As a result, our proposed method achieves state-of-the-art performance in the cross-modal recipe retrieval task on the Recipe1M dataset. We make code and models publicly available.

arXiv:2103.13023

Can Vision Transformers Learn without Natural Images?

Can we complete pre-training of Vision Transformers (ViT) without natural images and human-annotated labels? Although a pre-trained ViT seems to heavily rely on a large-scale dataset and human-annotated labels, recent large-scale datasets contain several problems in terms of privacy violations, inadequate fairness protection, and labor-intensive annotation. In the present paper, we pre-train ViT without any image collections and annotation labor. We experimentally verify that our proposed framework partially outperforms sophisticated Self-Supervised Learning (SSL) methods like SimCLRv2 and MoCov2 without using any natural images in the pre-training phase. Moreover, although the ViT pre-trained without natural images produces some different visualizations from ImageNet pre-trained ViT, it can interpret natural image datasets to a large extent. For example, the performance rates on the CIFAR-10 dataset are as follows: our proposal 97.6 vs. SimCLRv2 97.4 vs. ImageNet 98.0.

arXiv:2103.12676공식 코드 · Papers with Code

Self-supervised representation learning from 12-lead ECG data

Clinical 12-lead electrocardiography (ECG) is one of the most widely encountered kinds of biosignals. Despite the increased availability of public ECG datasets, label scarcity remains a central challenge in the field. Self-supervised learning represents a promising way to alleviate this issue. In this work, we put forward the first comprehensive assessment of self-supervised representation learning from clinical 12-lead ECG data. To this end, we adapt state-of-the-art self-supervised methods based on instance discrimination and latent forecasting to the ECG domain. In a first step, we learn contrastive representations and evaluate their quality based on linear evaluation performance on a recently established, comprehensive, clinical ECG classification task. In a second step, we analyze the impact of self-supervised pretraining on finetuned ECG classifiers as compared to purely supervised performance. For the best-performing method, an adaptation of contrastive predictive coding, we find a linear evaluation performance only 0.5% below supervised performance. For the finetuned models, we find improvements in downstream performance of roughly 1% compared to supervised performance, label efficiency, as well as robustness against physiological noise. This work clearly establishes the feasibility of extracting discriminative representations from ECG data via self-supervised learning and the numerous advantages when finetuning such representations on downstream tasks as compared to purely supervised training. As first comprehensive assessment of its kind in the ECG domain carried out exclusively on publicly available datasets, we hope to establish a first step towards reproducible progress in the rapidly evolving field of representation learning for biosignals.

arXiv:2103.12608공식 코드 · Papers with Code

DIG: A Turnkey Library for Diving into Graph Deep Learning Research

Although there exist several libraries for deep learning on graphs, they are aiming at implementing basic operations for graph deep learning. In the research community, implementing and benchmarking various advanced tasks are still painful and time-consuming with existing libraries. To facilitate graph deep learning research, we introduce DIG: Dive into Graphs, a turnkey library that provides a unified testbed for higher level, research-oriented graph deep learning tasks. Currently, we consider graph generation, self-supervised learning on graphs, explainability of graph neural networks, and deep learning on 3D graphs. For each direction, we provide unified implementations of data interfaces, common algorithms, and evaluation metrics. Altogether, DIG is an extensible, open-source, and turnkey library for researchers to develop new methods and effortlessly compare with common baselines using widely used datasets and evaluation metrics. Source code is available at https://github.com/divelab/DIG.

arXiv:2103.12496공식 코드 · Papers with Code

Revisiting Self-Supervised Monocular Depth Estimation

Self-supervised learning of depth map prediction and motion estimation from monocular video sequences is of vital importance -- since it realizes a broad range of tasks in robotics and autonomous vehicles. A large number of research efforts have enhanced the performance by tackling illumination variation, occlusions, and dynamic objects, to name a few. However, each of those efforts targets individual goals and endures as separate works. Moreover, most of previous works have adopted the same CNN architecture, not reaping architectural benefits. Therefore, the need to investigate the inter-dependency of the previous methods and the effect of architectural factors remains. To achieve these objectives, we revisit numerous previously proposed self-supervised methods for joint learning of depth and motion, perform a comprehensive empirical study, and unveil multiple crucial insights. Furthermore, we remarkably enhance the performance as a result of our study -- outperforming previous state-of-the-art performance.

arXiv:2103.12719

Characterizing and Improving the Robustness of Self-Supervised Learning through Background Augmentations

Recent progress in self-supervised learning has demonstrated promising results in multiple visual tasks. An important ingredient in high-performing self-supervised methods is the use of data augmentation by training models to place different augmented views of the same image nearby in embedding space. However, commonly used augmentation pipelines treat images holistically, ignoring the semantic relevance of parts of an image-e.g. a subject vs. a background-which can lead to the learning of spurious correlations. Our work addresses this problem by investigating a class of simple, yet highly effective "background augmentations", which encourage models to focus on semantically-relevant content by discouraging them from focusing on image backgrounds. Through a systematic investigation, we show that background augmentations lead to substantial improvements in performance across a spectrum of state-of-the-art self-supervised methods (MoCo-v2, BYOL, SwAV) on a variety of tasks, e.g. $\sim$+1-2% gains on ImageNet, enabling performance on par with the supervised baseline. Further, we find the improvement in limited-labels settings is even larger (up to 4.2%). Background augmentations also improve robustness to a number of distribution shifts, including natural adversarial examples, ImageNet-9, adversarial attacks, ImageNet-Renditions. We also make progress in completely unsupervised saliency detection, in the process of generating saliency masks used for background augmentations.

arXiv:2103.11031

Bootstrapped Self-Supervised Training with Monocular Video for Semantic Segmentation and Depth Estimation

For a robot deployed in the world, it is desirable to have the ability of autonomous learning to improve its initial pre-set knowledge. We formalize this as a bootstrapped self-supervised learning problem where a system is initially bootstrapped with supervised training on a labeled dataset and we look for a self-supervised training method that can subsequently improve the system over the supervised training baseline using only unlabeled data. In this work, we leverage temporal consistency between frames in monocular video to perform this bootstrapped self-supervised training. We show that a well-trained state-of-the-art semantic segmentation network can be further improved through our method. In addition, we show that the bootstrapped self-supervised training framework can help a network learn depth estimation better than pure supervised training or self-supervised training.

arXiv:2103.11275공식 코드 · Papers with Code

Self-supervised Representation Learning with Relative Predictive Coding

This paper introduces Relative Predictive Coding (RPC), a new contrastive representation learning objective that maintains a good balance among training stability, minibatch size sensitivity, and downstream task performance. The key to the success of RPC is two-fold. First, RPC introduces the relative parameters to regularize the objective for boundedness and low variance. Second, RPC contains no logarithm and exponential score functions, which are the main cause of training instability in prior contrastive objectives. We empirically verify the effectiveness of RPC on benchmark vision and speech self-supervised learning tasks. Lastly, we relate RPC with mutual information (MI) estimation, showing RPC can be used to estimate MI with low variance.

arXiv:2103.11703공식 코드 · Papers with Code

Model-based 3D Hand Reconstruction via Self-Supervised Learning

Reconstructing a 3D hand from a single-view RGB image is challenging due to various hand configurations and depth ambiguity. To reliably reconstruct a 3D hand from a monocular image, most state-of-the-art methods heavily rely on 3D annotations at the training stage, but obtaining 3D annotations is expensive. To alleviate reliance on labeled training data, we propose S2HAND, a self-supervised 3D hand reconstruction network that can jointly estimate pose, shape, texture, and the camera viewpoint. Specifically, we obtain geometric cues from the input image through easily accessible 2D detected keypoints. To learn an accurate hand reconstruction model from these noisy geometric cues, we utilize the consistency between 2D and 3D representations and propose a set of novel losses to rationalize outputs of the neural network. For the first time, we demonstrate the feasibility of training an accurate 3D hand reconstruction network without relying on manual annotations. Our experiments show that the proposed method achieves comparable performance with recent fully-supervised methods while using fewer supervision data.

arXiv:2103.10502

Ano-Graph: Learning Normal Scene Contextual Graphs to Detect Video Anomalies

Video anomaly detection has proved to be a challenging task owing to its unsupervised training procedure and high spatio-temporal complexity existing in real-world scenarios. In the absence of anomalous training samples, state-of-the-art methods try to extract features that fully grasp normal behaviors in both space and time domains using different approaches such as autoencoders, or generative adversarial networks. However, these approaches completely ignore or, by using the ability of deep networks in the hierarchical modeling, poorly model the spatio-temporal interactions that exist between objects. To address this issue, we propose a novel yet efficient method named Ano-Graph for learning and modeling the interaction of normal objects. Towards this end, a Spatio-Temporal Graph (STG) is made by considering each node as an object's feature extracted from a real-time off-the-shelf object detector, and edges are made based on their interactions. After that, a self-supervised learning method is employed on the STG in such a way that encapsulates interactions in a semantic space. Our method is data-efficient, significantly more robust against common real-world variations such as illumination, and passes SOTA by a large margin on the challenging datasets ADOC and Street Scene while stays competitive on Avenue, ShanghaiTech, and UCSD.

arXiv:2103.10056

Dementia Severity Classification under Small Sample Size and Weak Supervision in Thick Slice MRI

Early detection of dementia through specific biomarkers in MR images plays a critical role in developing support strategies proactively. Fazekas scale facilitates an accurate quantitative assessment of the severity of white matter lesions and hence the disease. Imaging Biomarkers of dementia are multiple and comprehensive documentation of them is time-consuming. Therefore, any effort to automatically extract these biomarkers will be of clinical value while reducing inter-rater discrepancies. To tackle this problem, we propose to classify the disease severity based on the Fazekas scale through the visual biomarkers, namely the Periventricular White Matter (PVWM) and the Deep White Matter (DWM) changes, in the real-world setting of thick-slice MRI. Small training sample size and weak supervision in form of assigning severity labels to the whole MRI stack are among the main challenges. To combat the mentioned issues, we have developed a deep learning pipeline that employs self-supervised representation learning, multiple instance learning, and appropriate pre-processing steps. We use pretext tasks such as non-linear transformation, local shuffling, in- and out-painting for self-supervised learning of useful features in this domain. Furthermore, an attention model is used to determine the relevance of each MRI slice for predicting the Fazekas scale in an unsupervised manner. We show the significant superiority of our method in distinguishing different classes of dementia compared to state-of-the-art methods in our mentioned setting, which improves the macro averaged F1-score of state-of-the-art from 61% to 76% in PVWM, and from 58% to 69.2% in DWM.