PyTorchKR Papers

최신 논문

총 2,023편
태그·기간 등 상세 필터

카테고리 탐색

로그인 후 이 태그 구독

arXiv:2210.01917공식 코드 · Papers with Code

Differentiable Raycasting for Self-supervised Occupancy Forecasting

Motion planning for safe autonomous driving requires learning how the environment around an ego-vehicle evolves with time. Ego-centric perception of driveable regions in a scene not only changes with the motion of actors in the environment, but also with the movement of the ego-vehicle itself. Self-supervised representations proposed for large-scale planning, such as ego-centric freespace, confound these two motions, making the representation difficult to use for downstream motion planners. In this paper, we use geometric occupancy as a natural alternative to view-dependent representations such as freespace. Occupancy maps naturally disentangle the motion of the environment from the motion of the ego-vehicle. However, one cannot directly observe the full 3D occupancy of a scene (due to occlusion), making it difficult to use as a signal for learning. Our key insight is to use differentiable raycasting to "render" future occupancy predictions into future LiDAR sweep predictions, which can be compared with ground-truth sweeps for self-supervised learning. The use of differentiable raycasting allows occupancy to emerge as an internal representation within the forecasting network. In the absence of groundtruth occupancy, we quantitatively evaluate the forecasting of raycasted LiDAR sweeps and show improvements of upto 15 F1 points. For downstream motion planners, where emergent occupancy can be directly used to guide non-driveable regions, this representation relatively reduces the number of collisions with objects by up to 17% as compared to freespace-centric motion planners.

arXiv:2210.02349공식 코드 · Papers with Code

Fitting a Directional Microstructure Model to Diffusion-Relaxation MRI Data with Self-Supervised Machine Learning

Machine learning is a powerful approach for fitting microstructural models to diffusion MRI data. Early machine learning microstructure imaging implementations trained regressors to estimate model parameters in a supervised way, using synthetic training data with known ground truth. However, a drawback of this approach is that the choice of training data impacts fitted parameter values. Self-supervised learning is emerging as an attractive alternative to supervised learning in this context. Thus far, both supervised and self-supervised learning have typically been applied to isotropic models, such as intravoxel incoherent motion (IVIM), as opposed to models where the directionality of anisotropic structures is also estimated. In this paper, we demonstrate self-supervised machine learning model fitting for a directional microstructural model. In particular, we fit a combined T1-ball-stick model to the multidimensional diffusion (MUDI) challenge diffusion-relaxation dataset. Our self-supervised approach shows clear improvements in parameter estimation and computational time, for both simulated and in-vivo brain data, compared to standard non-linear least squares fitting. Code for the artificial neural net constructed for this study is available for public use from the following GitHub repository: https://github.com/jplte/deep-T1-ball-stick

arXiv:2210.01696공식 코드 · Papers with Code

Clean self-supervised MRI reconstruction from noisy, sub-sampled training data with Robust SSDU

Most existing methods for Magnetic Resonance Imaging (MRI) reconstruction with deep learning use fully supervised training, which assumes that a high signal-to-noise ratio (SNR), fully sampled dataset is available for training. In many circumstances, however, such a dataset is highly impractical or even technically infeasible to acquire. Recently, a number of self-supervised methods for MR reconstruction have been proposed, which use sub-sampled data only. However, the majority of such methods, such as Self-Supervised Learning via Data Undersampling (SSDU), are susceptible to reconstruction errors arising from noise in the measured data. In response, we propose Robust SSDU, which provably recovers clean images from noisy, sub-sampled training data by simultaneously estimating missing k-space samples and denoising the available samples. Robust SSDU trains the reconstruction network to map from a further noisy and sub-sampled version of the data to the original, singly noisy and sub-sampled data, and applies an additive Noisier2Noise correction term at inference. We also present a related method, Noiser2Full, that recovers clean images when noisy, fully sampled data is available for training. Both proposed methods are applicable to any network architecture, straight-forward to implement and have similar computational cost to standard training. We evaluate our methods on the multi-coil fastMRI brain dataset with a novel denoising-specific architecture and find that it performs competitively with a benchmark trained on clean, fully sampled data.

arXiv:2210.01273공식 코드 · Papers with Code

An attention-based backend allowing efficient fine-tuning of transformer models for speaker verification

In recent years, self-supervised learning paradigm has received extensive attention due to its great success in various down-stream tasks. However, the fine-tuning strategies for adapting those pre-trained models to speaker verification task have yet to be fully explored. In this paper, we analyze several feature extraction approaches built on top of a pre-trained model, as well as regularization and learning rate schedule to stabilize the fine-tuning process and further boost performance: multi-head factorized attentive pooling is proposed to factorize the comparison of speaker representations into multiple phonetic clusters. We regularize towards the parameters of the pre-trained model and we set different learning rates for each layer of the pre-trained model during fine-tuning. The experimental results show our method can significantly shorten the training time to 4 hours and achieve SOTA performance: 0.59%, 0.79% and 1.77% EER on Vox1-O, Vox1-E and Vox1-H, respectively.

arXiv:2210.01703공식 코드 · Papers with Code

Improving Label-Deficient Keyword Spotting Through Self-Supervised Pretraining

Keyword Spotting (KWS) models are becoming increasingly integrated into various systems, e.g. voice assistants. To achieve satisfactory performance, these models typically rely on a large amount of labelled data, limiting their applications only to situations where such data is available. Self-supervised Learning (SSL) methods can mitigate such a reliance by leveraging readily-available unlabelled data. Most SSL methods for speech have primarily been studied for large models, whereas this is not ideal, as compact KWS models are generally required. This paper explores the effectiveness of SSL on small models for KWS and establishes that SSL can enhance the performance of small KWS models when labelled data is scarce. We pretrain three compact transformer-based KWS models using Data2Vec, and fine-tune them on a label-deficient setup of the Google Speech Commands data set. It is found that Data2Vec pretraining leads to a significant increase in accuracy, with label-deficient scenarios showing an improvement of 8.22% 11.18% absolute accuracy.

arXiv:2210.01571공식 코드 · Papers with Code

VICRegL: Self-Supervised Learning of Local Visual Features

Most recent self-supervised methods for learning image representations focus on either producing a global feature with invariance properties, or producing a set of local features. The former works best for classification tasks while the latter is best for detection and segmentation tasks. This paper explores the fundamental trade-off between learning local and global features. A new method called VICRegL is proposed that learns good global and local features simultaneously, yielding excellent performance on detection and segmentation tasks while maintaining good performance on classification tasks. Concretely, two identical branches of a standard convolutional net architecture are fed two differently distorted versions of the same image. The VICReg criterion is applied to pairs of global feature vectors. Simultaneously, the VICReg criterion is applied to pairs of local feature vectors occurring before the last pooling layer. Two local feature vectors are attracted to each other if their l2-distance is below a threshold or if their relative locations are consistent with a known geometric transformation between the two input images. We demonstrate strong performance on linear classification and segmentation transfer tasks. Code and pretrained models are publicly available at: https://github.com/facebookresearch/VICRegL

arXiv:2210.01116공식 코드 · Papers with Code

That Sounds Right: Auditory Self-Supervision for Dynamic Robot Manipulation

Learning to produce contact-rich, dynamic behaviors from raw sensory data has been a longstanding challenge in robotics. Prominent approaches primarily focus on using visual or tactile sensing, where unfortunately one fails to capture high-frequency interaction, while the other can be too delicate for large-scale data collection. In this work, we propose a data-centric approach to dynamic manipulation that uses an often ignored source of information: sound. We first collect a dataset of 25k interaction-sound pairs across five dynamic tasks using commodity contact microphones. Then, given this data, we leverage self-supervised learning to accelerate behavior prediction from sound. Our experiments indicate that this self-supervised 'pretraining' is crucial to achieving high performance, with a 34.5% lower MSE than plain supervised learning and a 54.3% lower MSE over visual training. Importantly, we find that when asked to generate desired sound profiles, online rollouts of our models on a UR10 robot can produce dynamic behavior that achieves an average of 11.5% improvement over supervised learning on audio similarity metrics.

arXiv:2210.00825공식 코드 · Papers with Code

Self-omics: A Self-supervised Learning Framework for Multi-omics Cancer Data

We have gained access to vast amounts of multi-omics data thanks to Next Generation Sequencing. However, it is challenging to analyse this data due to its high dimensionality and much of it not being annotated. Lack of annotated data is a significant problem in machine learning, and Self-Supervised Learning (SSL) methods are typically used to deal with limited labelled data. However, there is a lack of studies that use SSL methods to exploit inter-omics relationships on unlabelled multi-omics data. In this work, we develop a novel and efficient pre-training paradigm that consists of various SSL components, including but not limited to contrastive alignment, data recovery from corrupted samples, and using one type of omics data to recover other omic types. Our pre-training paradigm improves performance on downstream tasks with limited labelled data. We show that our approach outperforms the state-of-the-art method in cancer type classification on the TCGA pan-cancer dataset in semi-supervised setting. Moreover, we show that the encoders that are pre-trained using our approach can be used as powerful feature extractors even without fine-tuning. Our ablation study shows that the method is not overly dependent on any pretext task component. The network architectures in our approach are designed to handle missing omic types and multiple datasets for pre-training and downstream training. Our pre-training paradigm can be extended to perform zero-shot classification of rare cancers.

arXiv:2210.00643공식 코드 · Papers with Code

Spectral Augmentation for Self-Supervised Learning on Graphs

Graph contrastive learning (GCL), as an emerging self-supervised learning technique on graphs, aims to learn representations via instance discrimination. Its performance heavily relies on graph augmentation to reflect invariant patterns that are robust to small perturbations; yet it still remains unclear about what graph invariance GCL should capture. Recent studies mainly perform topology augmentations in a uniformly random manner in the spatial domain, ignoring its influence on the intrinsic structural properties embedded in the spectral domain. In this work, we aim to find a principled way for topology augmentations by exploring the invariance of graphs from the spectral perspective. We develop spectral augmentation which guides topology augmentations by maximizing the spectral change. Extensive experiments on both graph and node classification tasks demonstrate the effectiveness of our method in self-supervised representation learning. The proposed method also brings promising generalization capability in transfer learning, and is equipped with intriguing robustness property under adversarial attacks. Our study sheds light on a general principle for graph topology augmentation.

arXiv:2209.15525공식 코드 · Papers with Code

Slimmable Networks for Contrastive Self-supervised Learning

Self-supervised learning makes significant progress in pre-training large models, but struggles with small models. Mainstream solutions to this problem rely mainly on knowledge distillation, which involves a two-stage procedure: first training a large teacher model and then distilling it to improve the generalization ability of smaller ones. In this work, we introduce another one-stage solution to obtain pre-trained small models without the need for extra teachers, namely, slimmable networks for contrastive self-supervised learning (SlimCLR). A slimmable network consists of a full network and several weight-sharing sub-networks, which can be pre-trained once to obtain various networks, including small ones with low computation costs. However, interference between weight-sharing networks leads to severe performance degradation in self-supervised cases, as evidenced by gradient magnitude imbalance and gradient direction divergence. The former indicates that a small proportion of parameters produce dominant gradients during backpropagation, while the main parameters may not be fully optimized. The latter shows that the gradient direction is disordered, and the optimization process is unstable. To address these issues, we introduce three techniques to make the main parameters produce dominant gradients and sub-networks have consistent outputs. These techniques include slow start training of sub-networks, online distillation, and loss re-weighting according to model sizes. Furthermore, theoretical results are presented to demonstrate that a single slimmable linear layer is sub-optimal during linear evaluation. Thus a switchable linear probe layer is applied during linear evaluation. We instantiate SlimCLR with typical contrastive learning frameworks and achieve better performance than previous arts with fewer parameters and FLOPs. The code is at https://github.com/mzhaoshuai/SlimCLR.

arXiv:2209.14345공식 코드 · Papers with Code

Audio Barlow Twins: Self-Supervised Audio Representation Learning

The Barlow Twins self-supervised learning objective requires neither negative samples or asymmetric learning updates, achieving results on a par with the current state-of-the-art within Computer Vision. As such, we present Audio Barlow Twins, a novel self-supervised audio representation learning approach, adapting Barlow Twins to the audio domain. We pre-train on the large-scale audio dataset AudioSet, and evaluate the quality of the learnt representations on 18 tasks from the HEAR 2021 Challenge, achieving results which outperform, or otherwise are on a par with, the current state-of-the-art for instance discrimination self-supervised learning approaches to audio representation learning. Code at https://github.com/jonahanton/SSL_audio.

arXiv:2209.12900공식 코드 · Papers with Code

The Efficacy of Self-Supervised Speech Models for Audio Representations

Self-supervised learning (SSL) speech models, which can serve as powerful upstream models to extract meaningful speech representations, have achieved unprecedented success in speech representation learning. However, their effectiveness on non-speech datasets is relatively less explored. In this work, we propose an ensemble framework, with a combination of ensemble techniques, to fuse SSL speech models' embeddings. Extensive experiments on speech and non-speech audio datasets are conducted to investigate the representation abilities of our ensemble method and its single constituent model. Ablation studies are carried out to evaluate the performances of different ensemble techniques, such as feature averaging and concatenation. All experiments are conducted during NeurIPS 2021 HEAR Challenge as a standard evaluation pipeline provided by competition officials. Results demonstrate SSL speech models' strong abilities on various non-speech tasks, while we also note that they fail to deal with fine-grained music tasks, such as pitch classification and note onset detection. In addition, feature ensemble is shown to have great potential on producing more holistic representations, as our proposed framework generally surpasses state-of-the-art SSL speech/audio models and has superior performance on various datasets compared with other teams in HEAR Challenge. Our code is available at https://github.com/tony10101105/HEAR-2021-NeurIPS-Challenge -- NTU-GURA.

arXiv:2209.12157공식 코드 · Papers with Code

Dive into Self-Supervised Learning for Medical Image Analysis: Data, Models and Tasks

Self-supervised learning (SSL) has achieved remarkable performance in various medical imaging tasks by dint of priors from massive unlabelled data. However, regarding a specific downstream task, there is still a lack of an instruction book on how to select suitable pretext tasks and implementation details throughout the standard ``pretrain-then-finetune'' workflow. In this work, we focus on exploiting the capacity of SSL in terms of four realistic and significant issues: (1) the impact of SSL on imbalanced datasets, (2) the network architecture, (3) the applicability of upstream tasks to downstream tasks and (4) the stacking effect of SSL and common policies for deep learning. We provide a large-scale, in-depth and fine-grained study through extensive experiments on predictive, contrastive, generative and multi-SSL algorithms. Based on the results, we have uncovered several insights. Positively, SSL advances class-imbalanced learning mainly by boosting the performance of the rare class, which is of interest to clinical diagnosis. Unfortunately, SSL offers marginal or even negative returns in some cases, including severely imbalanced and relatively balanced data regimes, as well as combinations with common training policies. Our intriguing findings provide practical guidelines for the usage of SSL in the medical context and highlight the need for developing universal pretext tasks to accommodate diverse application scenarios.

arXiv:2209.12621공식 코드 · Papers with Code

Improving Image Clustering through Sample Ranking and Its Application to remote--sensing images

Image clustering is a very useful technique that is widely applied to various areas, including remote sensing. Recently, visual representations by self-supervised learning have greatly improved the performance of image clustering. To further improve the well-trained clustering models, this paper proposes a novel method by first ranking samples within each cluster based on the confidence in their belonging to the current cluster and then using the ranking to formulate a weighted cross-entropy loss to train the model. For ranking the samples, we developed a method for computing the likelihood of samples belonging to the current clusters based on whether they are situated in densely populated neighborhoods, while for training the model, we give a strategy for weighting the ranked samples. We present extensive experimental results that demonstrate that the new technique can be used to improve the State-of-the-Art image clustering models, achieving accuracy performance gains ranging from $2.1\%$ to $15.9\%$. Performing our method on a variety of datasets from remote sensing, we show that our method can be effectively applied to remote--sensing images.

arXiv:2209.11359공식 코드 · Papers with Code

CUTS: A Deep Learning and Topological Framework for Multigranular Unsupervised Medical Image Segmentation

Segmenting medical images is critical to facilitating both patient diagnoses and quantitative research. A major limiting factor is the lack of labeled data, as obtaining expert annotations for each new set of imaging data and task can be labor intensive and inconsistent among annotators. We present CUTS, an unsupervised deep learning framework for medical image segmentation. CUTS operates in two stages. For each image, it produces an embedding map via intra-image contrastive learning and local patch reconstruction. Then, these embeddings are partitioned at dynamic granularity levels that correspond to the data topology. CUTS yields a series of coarse-to-fine-grained segmentations that highlight features at various granularities. We applied CUTS to retinal fundus images and two types of brain MRI images to delineate structures and patterns at different scales. When evaluated against predefined anatomical masks, CUTS improved the dice coefficient and Hausdorff distance by at least 10% compared to existing unsupervised methods. Finally, CUTS showed performance on par with Segment Anything Models (SAM, MedSAM, SAM-Med2D) pre-trained on gigantic labeled datasets.

arXiv:2209.10901공식 코드 · Papers with Code

Pretraining the Vision Transformer using self-supervised methods for vision based Deep Reinforcement Learning

The Vision Transformer architecture has shown to be competitive in the computer vision (CV) space where it has dethroned convolution-based networks in several benchmarks. Nevertheless, convolutional neural networks (CNN) remain the preferential architecture for the representation module in reinforcement learning. In this work, we study pretraining a Vision Transformer using several state-of-the-art self-supervised methods and assess the quality of the learned representations. To show the importance of the temporal dimension in this context we propose an extension of VICReg to better capture temporal relations between observations by adding a temporal order verification task. Our results show that all methods are effective in learning useful representations and avoiding representational collapse for observations from Atari Learning Environment (ALE) which leads to improvements in data efficiency when we evaluated in reinforcement learning (RL). Moreover, the encoder pretrained with the temporal order verification task shows the best results across all experiments, with richer representations, more focused attention maps and sparser representation vectors throughout the layers of the encoder, which shows the importance of exploring such similarity dimension. With this work, we hope to provide some insights into the representations learned by ViT during a self-supervised pretraining with observations from RL environments and which properties arise in the representations that lead to the best-performing agents. The source code will be available at: https://github.com/mgoulao/TOV-VICReg

arXiv:2209.09019공식 코드 · Papers with Code

LAVIS: A Library for Language-Vision Intelligence

We introduce LAVIS, an open-source deep learning library for LAnguage-VISion research and applications. LAVIS aims to serve as a one-stop comprehensive library that brings recent advancements in the language-vision field accessible for researchers and practitioners, as well as fertilizing future research and development. It features a unified interface to easily access state-of-the-art image-language, video-language models and common datasets. LAVIS supports training, evaluation and benchmarking on a rich variety of tasks, including multimodal classification, retrieval, captioning, visual question answering, dialogue and pre-training. In the meantime, the library is also highly extensible and configurable, facilitating future development and customization. In this technical report, we describe design principles, key components and functionalities of the library, and also present benchmarking results across common language-vision tasks. The library is available at: https://github.com/salesforce/LAVIS.

arXiv:2209.08887공식 코드 · Papers with Code

Attentive Symmetric Autoencoder for Brain MRI Segmentation

Self-supervised learning methods based on image patch reconstruction have witnessed great success in training auto-encoders, whose pre-trained weights can be transferred to fine-tune other downstream tasks of image understanding. However, existing methods seldom study the various importance of reconstructed patches and the symmetry of anatomical structures, when they are applied to 3D medical images. In this paper we propose a novel Attentive Symmetric Auto-encoder (ASA) based on Vision Transformer (ViT) for 3D brain MRI segmentation tasks. We conjecture that forcing the auto-encoder to recover informative image regions can harvest more discriminative representations, than to recover smooth image patches. Then we adopt a gradient based metric to estimate the importance of each image patch. In the pre-training stage, the proposed auto-encoder pays more attention to reconstruct the informative patches according to the gradient metrics. Moreover, we resort to the prior of brain structures and develop a Symmetric Position Encoding (SPE) method to better exploit the correlations between long-range but spatially symmetric regions to obtain effective features. Experimental results show that our proposed attentive symmetric auto-encoder outperforms the state-of-the-art self-supervised learning methods and medical image segmentation models on three brain MRI segmentation benchmarks.

arXiv:2209.08776공식 코드 · Papers with Code

NeRF-SOS: Any-View Self-supervised Object Segmentation on Complex Scenes

Neural volumetric representations have shown the potential that Multi-layer Perceptrons (MLPs) can be optimized with multi-view calibrated images to represent scene geometry and appearance, without explicit 3D supervision. Object segmentation can enrich many downstream applications based on the learned radiance field. However, introducing hand-crafted segmentation to define regions of interest in a complex real-world scene is non-trivial and expensive as it acquires per view annotation. This paper carries out the exploration of self-supervised learning for object segmentation using NeRF for complex real-world scenes. Our framework, called NeRF with Self-supervised Object Segmentation NeRF-SOS, couples object segmentation and neural radiance field to segment objects in any view within a scene. By proposing a novel collaborative contrastive loss in both appearance and geometry levels, NeRF-SOS encourages NeRF models to distill compact geometry-aware segmentation clusters from their density fields and the self-supervised pre-trained 2D visual features. The self-supervised object segmentation framework can be applied to various NeRF models that both lead to photo-realistic rendering results and convincing segmentation maps for both indoor and outdoor scenarios. Extensive results on the LLFF, Tank & Temple, and BlendedMVS datasets validate the effectiveness of NeRF-SOS. It consistently surpasses other 2D-based self-supervised baselines and predicts finer semantics masks than existing supervised counterparts. Please refer to the video on our project page for more details:https://zhiwenfan.github.io/NeRF-SOS.

arXiv:2209.07629공식 코드 · Papers with Code

Self-Relation Attention and Temporal Awareness for Emotion Recognition via Vocal Burst

The technical report presents our emotion recognition pipeline for high-dimensional emotion task (A-VB High) in The ACII Affective Vocal Bursts (A-VB) 2022 Workshop \& Competition. Our proposed method contains three stages. Firstly, we extract the latent features from the raw audio signal and its Mel-spectrogram by self-supervised learning methods. Then, the features from the raw signal are fed to the self-relation attention and temporal awareness (SA-TA) module for learning the valuable information between these latent features. Finally, we concatenate all the features and utilize a fully-connected layer to predict each emotion's score. By empirical experiments, our proposed method achieves a mean concordance correlation coefficient (CCC) of 0.7295 on the test set, compared to 0.5686 on the baseline model. The code of our method is available at https://github.com/linhtd812/A-VB2022.