Regression that predicts continuous quantity is a central part of applications using computational imaging and computer vision technologies. Yet, studying and understanding self-supervised learning for regression tasks - except for a particular regression task, image denoising - have lagged behind. This paper proposes a general self-supervised regression learning (SSRL) framework that enables learning regression neural networks with only input data (but without ground-truth target data), by using a designable pseudo-predictor that encapsulates domain knowledge of a specific application. The paper underlines the importance of using domain knowledge by showing that under different settings, the better pseudo-predictor can lead properties of SSRL closer to those of ordinary supervised learning. Numerical experiments for low-dose computational tomography denoising and camera image denoising demonstrate that proposed SSRL significantly improves the denoising quality over several existing self-supervised denoising methods.
The prior self-supervised learning researches mainly select image-level instance discrimination as pretext task. It achieves a fantastic classification performance that is comparable to supervised learning methods. However, with degraded transfer performance on downstream tasks such as object detection. To bridge the performance gap, we propose a novel object-level self-supervised learning method, called Contrastive learning with Downstream background invariance (CoDo). The pretext task is converted to focus on instance location modeling for various backgrounds, especially for downstream datasets. The ability of background invariance is considered vital for object detection. Firstly, a data augmentation strategy is proposed to paste the instances onto background images, and then jitter the bounding box to involve background information. Secondly, we implement architecture alignment between our pretraining network and the mainstream detection pipelines. Thirdly, hierarchical and multi views contrastive learning is designed to improve performance of visual representation learning. Experiments on MSCOCO demonstrate that the proposed CoDo with common backbones, ResNet50-FPN, yields strong transfer learning results for object detection.
Self-Supervised Learning (SSL) has made great strides recently. SSL speech models achieve decent performance on a wide range of downstream tasks, suggesting that they extract different aspects of information from speech. However, how SSL models store various information in hidden representations without interfering is still poorly understood. Taking the recently successful SSL model, HuBERT, as an example, we explore how the SSL model processes and stores speaker information in the representation. We found that HuBERT stores speaker information in representations whose positions correspond to silences in a waveform. There are several pieces of evidence. (1) We find that the utterances with more silent parts in the waveforms have better Speaker Identification (SID) accuracy. (2) If we use the whole utterances for SID, the silence part always contributes more to the SID task. (3) If we only use the representation of a part of the utterance for SID, the silenced part has higher accuracy than the other parts. Our findings not only contribute to a better understanding of SSL models but also improve performance. By simply adding silence to the original waveform, HuBERT improved its accuracy on SID by nearly 2%.
Enormous efforts have been recently made to super-resolve hyperspectral (HS) images with the aid of high spatial resolution multispectral (MS) images. Most prior works usually perform the fusion task by means of multifarious pixel-level priors. Yet the intrinsic effects of a large distribution gap between HS-MS data due to differences in the spatial and spectral resolution are less investigated. The gap might be caused by unknown sensor-specific properties or highly-mixed spectral information within one pixel (due to low spatial resolution). To this end, we propose a subpixel-level HS super-resolution framework by devising a novel decoupled-and-coupled network, called DC-Net, to progressively fuse HS-MS information from the pixel- to subpixel-level, from the image- to feature-level. As the name suggests, DC-Net first decouples the input into common (or cross-sensor) and sensor-specific components to eliminate the gap between HS-MS images before further fusion, and then fully blends them by a model-guided coupled spectral unmixing (CSU) net. More significantly, we append a self-supervised learning module behind the CSU net by guaranteeing the material consistency to enhance the detailed appearances of the restored HS product. Extensive experimental results show the superiority of our method both visually and quantitatively and achieve a significant improvement in comparison with the state-of-the-arts. Furthermore, the codes and datasets will be available at https://sites.google.com/view/danfeng-hong for the sake of reproducibility.
Deep learning methods have been successfully used in various computer vision tasks. Inspired by that success, deep learning has been explored in magnetic resonance imaging (MRI) reconstruction. In particular, integrating deep learning and model-based optimization methods has shown considerable advantages. However, a large amount of labeled training data is typically needed for high reconstruction quality, which is challenging for some MRI applications. In this paper, we propose a novel reconstruction method, named DURED-Net, that enables interpretable self-supervised learning for MR image reconstruction by combining a self-supervised denoising network and a plug-and-play method. We aim to boost the reconstruction performance of Noise2Noise in MR reconstruction by adding an explicit prior that utilizes imaging physics. Specifically, the leverage of a denoising network for MRI reconstruction is achieved using Regularization by Denoising (RED). Experiment results demonstrate that the proposed method requires a reduced amount of training data to achieve high reconstruction quality among the state-of-art of MR reconstruction utilizing the Noise2Noise method.
Large numbers of labeled medical images are essential for the accurate detection of anomalies, but manual annotation is labor-intensive and time-consuming. Self-supervised learning (SSL) is a training method to learn data-specific features without manual annotation. Several SSL-based models have been employed in medical image anomaly detection. These SSL methods effectively learn representations in several field-specific images, such as natural and industrial product images. However, owing to the requirement of medical expertise, typical SSL-based models are inefficient in medical image anomaly detection. We present an SSL-based model that enables anatomical structure-based unsupervised anomaly detection (UAD). The model employs the anatomy-aware pasting (AnatPaste) augmentation tool. AnatPaste employs a threshold-based lung segmentation pretext task to create anomalies in normal chest radiographs, which are used for model pretraining. These anomalies are similar to real anomalies and help the model recognize them. We evaluate our model on three opensource chest radiograph datasets. Our model exhibit area under curves (AUC) of 92.1%, 78.7%, and 81.9%, which are the highest among existing UAD models. This is the first SSL model to employ anatomical information as a pretext task. AnatPaste can be applied in various deep learning models and downstream tasks. It can be employed for other modalities by fixing appropriate segmentation. Our code is publicly available at: https://github.com/jun-sato/AnatPaste.
Incremental few-shot object detection aims at detecting novel classes without forgetting knowledge of the base classes with only a few labeled training data from the novel classes. Most related prior works are on incremental object detection that rely on the availability of abundant training samples per novel class that substantially limits the scalability to real-world setting where novel data can be scarce. In this paper, we propose the Incremental-DETR that does incremental few-shot object detection via fine-tuning and self-supervised learning on the DETR object detector. To alleviate severe over-fitting with few novel class data, we first fine-tune the class-specific components of DETR with self-supervision from additional object proposals generated using Selective Search as pseudo labels. We further introduce an incremental few-shot fine-tuning strategy with knowledge distillation on the class-specific components of DETR to encourage the network in detecting novel classes without forgetting the base classes. Extensive experiments conducted on standard incremental object detection and incremental few-shot object detection settings show that our approach significantly outperforms state-of-the-art methods by a large margin.
Until recently, the number of public real-world text images was insufficient for training scene text recognizers. Therefore, most modern training methods rely on synthetic data and operate in a fully supervised manner. Nevertheless, the amount of public real-world text images has increased significantly lately, including a great deal of unlabeled data. Leveraging these resources requires semi-supervised approaches; however, the few existing methods do not account for vision-language multimodality structure and therefore suboptimal for state-of-the-art multimodal architectures. To bridge this gap, we present semi-supervised learning for multimodal text recognizers (SemiMTR) that leverages unlabeled data at each modality training phase. Notably, our method refrains from extra training stages and maintains the current three-stage multimodal training procedure. Our algorithm starts by pretraining the vision model through a single-stage training that unifies self-supervised learning with supervised training. More specifically, we extend an existing visual representation learning algorithm and propose the first contrastive-based method for scene text recognition. After pretraining the language model on a text corpus, we fine-tune the entire network via a sequential, character-level, consistency regularization between weakly and strongly augmented views of text images. In a novel setup, consistency is enforced on each modality separately. Extensive experiments validate that our method outperforms the current training schemes and achieves state-of-the-art results on multiple scene text recognition benchmarks.
Self-supervised learning is a growing paradigm in deep representation learning, showing great generalization capabilities and competitive performance in low-labeled data regimes. The SimCLR framework proposes the NT-Xent loss for contrastive representation learning. The objective of the loss function is to maximize agreement, similarity, between sampled positive pairs. This short paper derives and proposes an upper bound for the loss and average similarity. An analysis of the implications is however not provided, but we strongly encourage anyone in the field to conduct this.
Stride length estimation using inertial measurement unit (IMU) sensors is getting popular recently as one representative gait parameter for health care and sports training. The traditional estimation method requires some explicit calibrations and design assumptions. Current deep learning methods suffer from few labeled data problem. To solve above problems, this paper proposes a single convolutional neural network (CNN) model to predict stride length of running and walking and classify the running or walking type per stride. The model trains its pretext task with self-supervised learning on a large unlabeled dataset for feature learning, and its downstream task on the stride length estimation and classification tasks with supervised learning with a small labeled dataset. The proposed model can achieve better average percent error, 4.78\%, on running and walking stride length regression and 99.83\% accuracy on running and walking classification, when compared to the previous approach, 7.44\% on the stride length estimation.
Machine learning (ML) models have been widely successful in the prediction of material properties. However, large labeled datasets required for training accurate ML models are elusive and computationally expensive to generate. Recent advances in Self-Supervised Learning (SSL) frameworks capable of training ML models on unlabeled data have mitigated this problem and demonstrated superior performance in computer vision and natural language processing tasks. Drawing inspiration from the developments in SSL, we introduce Crystal Twins (CT): an SSL method for crystalline materials property prediction. Using a large unlabeled dataset, we pre-train a Graph Neural Network (GNN) by applying the redundancy reduction principle to the graph latent embeddings of augmented instances obtained from the same crystalline system. By sharing the pre-trained weights when fine-tuning the GNN for regression tasks, we significantly improve the performance for 7 challenging material property prediction benchmarks
Self-Supervised learning (SSL) has become the new state-of-art in several domain classification and segmentation tasks. Of these, one popular category in SSL is distillation networks such as BYOL. This work proposes RSDnet, which applies the distillation network (BYOL) in the remote sensing (RS) domain where data is non-trivially different from natural RGB images. Since Multi-spectral (MS) and synthetic aperture radar (SAR) sensors provide varied spectral and spatial resolution information, we utilised them as an implicit augmentation to learn invariant feature embeddings. In order to learn RS based invariant features with SSL, we trained RSDnet in two ways, i.e., single channel feature learning and three channel feature learning. This work explores the usefulness of single channel feature learning from random MS and SAR bands compared to the common notion of using three or more bands. In our linear evaluation, these single channel features reached a 0.92 F1 score on the EuroSAT classification task and 59.6 mIoU on the DFC segmentation task for certain single bands. We also compared our results with ImageNet weights and showed that the RS based SSL model outperforms the supervised ImageNet based model. We further explored the usefulness of multi-modal data compared to single modality data, and it is shown that utilising MS and SAR data learn better invariant representations than utilising only MS data.
Recognizing transformation types applied to a video clip (RecogTrans) is a long-established paradigm for self-supervised video representation learning, which achieves much inferior performance compared to instance discrimination approaches (InstDisc) in recent works. However, based on a thorough comparison of representative RecogTrans and InstDisc methods, we observe the great potential of RecogTrans on both semantic-related and temporal-related downstream tasks. Based on hard-label classification, existing RecogTrans approaches suffer from noisy supervision signals in pre-training. To mitigate this problem, we developed TransRank, a unified framework for recognizing Transformations in a Ranking formulation. TransRank provides accurate supervision signals by recognizing transformations relatively, consistently outperforming the classification-based formulation. Meanwhile, the unified framework can be instantiated with an arbitrary set of temporal or spatial transformations, demonstrating good generality. With a ranking-based formulation and several empirical practices, we achieve competitive performance on video retrieval and action recognition. Under the same setting, TransRank surpasses the previous state-of-the-art method by 6.4% on UCF101 and 8.3% on HMDB51 for action recognition (Top1 Acc); improves video retrieval on UCF101 by 20.4% (R@1). The promising results validate that RecogTrans is still a worth exploring paradigm for video self-supervised learning. Codes will be released at https://github.com/kennymckormick/TransRank.
Definitive cancer diagnosis and management depend upon the extraction of information from microscopy images by pathologists. These images contain complex information requiring time-consuming expert human interpretation that is prone to human bias. Supervised deep learning approaches have proven powerful for classification tasks, but they are inherently limited by the cost and quality of annotations used for training these models. To address this limitation of supervised methods, we developed Histomorphological Phenotype Learning (HPL), a fully blue{self-}supervised methodology that requires no expert labels or annotations and operates via the automatic discovery of discriminatory image features in small image tiles. Tiles are grouped into morphologically similar clusters which constitute a library of histomorphological phenotypes, revealing trajectories from benign to malignant tissue via inflammatory and reactive phenotypes. These clusters have distinct features which can be identified using orthogonal methods, linking histologic, molecular and clinical phenotypes. Applied to lung cancer tissues, we show that they align closely with patient survival, with histopathologically recognised tumor types and growth patterns, and with transcriptomic measures of immunophenotype. We then demonstrate that these properties are maintained in a multi-cancer study. These results show the clusters represent recurrent host responses and modes of tumor growth emerging under natural selection. Code, pre-trained models, learned embeddings, and documentation are available to the community at https://github.com/AdalbertoCq/Histomorphological-Phenotype-Learning
In this paper, we introduce Masked Anomaly Detection (MAD), a general self-supervised learning task for multivariate time series anomaly detection. With the increasing availability of sensor data from industrial systems, being able to detecting anomalies from streams of multivariate time series data is of significant importance. Given the scarcity of anomalies in real-world applications, the majority of literature has been focusing on modeling normality. The learned normal representations can empower anomaly detection as the model has learned to capture certain key underlying data regularities. A typical formulation is to learn a predictive model, i.e., use a window of time series data to predict future data values. In this paper, we propose an alternative self-supervised learning task. By randomly masking a portion of the inputs and training a model to estimate them using the remaining ones, MAD is an improvement over the traditional left-to-right next step prediction (NSP) task. Our experimental results demonstrate that MAD can achieve better anomaly detection rates over traditional NSP approaches when using exactly the same neural network (NN) base models, and can be modified to run as fast as NSP models during test time on the same hardware, thus making it an ideal upgrade for many existing NSP-based NN anomaly detection models.
WiFi sensing technology has shown superiority in smart homes among various sensors for its cost-effective and privacy-preserving merits. It is empowered by Channel State Information (CSI) extracted from WiFi signals and advanced machine learning models to analyze motion patterns in CSI. Many learning-based models have been proposed for kinds of applications, but they severely suffer from environmental dependency. Though domain adaptation methods have been proposed to tackle this issue, it is not practical to collect high-quality, well-segmented and balanced CSI samples in a new environment for adaptation algorithms, but randomly-captured CSI samples can be easily collected. {\color{black}In this paper, we firstly explore how to learn a robust model from these low-quality CSI samples, and propose AutoFi, an annotation-efficient WiFi sensing model based on a novel geometric self-supervised learning algorithm.} The AutoFi fully utilizes unlabeled low-quality CSI samples that are captured randomly, and then transfers the knowledge to specific tasks defined by users, which is the first work to achieve cross-task transfer in WiFi sensing. The AutoFi is implemented on a pair of Atheros WiFi APs for evaluation. The AutoFi transfers knowledge from randomly collected CSI samples into human gait recognition and achieves state-of-the-art performance. Furthermore, we simulate cross-task transfer using public datasets to further demonstrate its capacity for cross-task learning. For the UT-HAR and Widar datasets, the AutoFi achieves satisfactory results on activity recognition and gesture recognition without any prior training. We believe that the AutoFi takes a huge step toward automatic WiFi sensing without any developer engagement.
Previous electroencephalogram (EEG) emotion recognition relies on single-task learning, which may lead to overfitting and learned emotion features lacking generalization. In this paper, a graph-based multi-task self-supervised learning model (GMSS) for EEG emotion recognition is proposed. GMSS has the ability to learn more general representations by integrating multiple self-supervised tasks, including spatial and frequency jigsaw puzzle tasks, and contrastive learning tasks. By learning from multiple tasks simultaneously, GMSS can find a representation that captures all of the tasks thereby decreasing the chance of overfitting on the original task, i.e., emotion recognition task. In particular, the spatial jigsaw puzzle task aims to capture the intrinsic spatial relationships of different brain regions. Considering the importance of frequency information in EEG emotional signals, the goal of the frequency jigsaw puzzle task is to explore the crucial frequency bands for EEG emotion recognition. To further regularize the learned features and encourage the network to learn inherent representations, contrastive learning task is adopted in this work by mapping the transformed data into a common feature space. The performance of the proposed GMSS is compared with several popular unsupervised and supervised methods. Experiments on SEED, SEED-IV, and MPED datasets show that the proposed model has remarkable advantages in learning more discriminative and general features for EEG emotional signals.
The simplified molecular-input line-entry system (SMILES) is the most popular representation of chemical compounds. Therefore, many SMILES-based molecular property prediction models have been developed. In particular, transformer-based models show promising performance because the model utilizes a massive chemical dataset for self-supervised learning. However, there is no transformer-based model to overcome the inherent limitations of SMILES, which result from the generation process of SMILES. In this study, we grammatically parsed SMILES to obtain connectivity between substructures and their type, which is called the grammatical knowledge of SMILES. First, we pretrained the transformers with substructural tokens, which were parsed from SMILES. Then, we used the training strategy 'same compound model' to better understand SMILES grammar. In addition, we injected knowledge of connectivity and type into the transformer with knowledge adapters. As a result, our representation model outperformed previous compound representations for the prediction of molecular properties. Finally, we analyzed the attention of the transformer model and adapters, demonstrating that the proposed model understands the grammar of SMILES.
State-of-the-art lidar-based 3D object detection methods rely on supervised learning and large labeled datasets. However, annotating lidar data is resource-consuming, and depending only on supervised learning limits the applicability of trained models. Self-supervised training strategies can alleviate these issues by learning a general point cloud backbone model for downstream 3D vision tasks. Against this backdrop, we show the relationship between self-supervised multi-frame flow representations and single-frame 3D detection hypotheses. Our main contribution leverages learned flow and motion representations and combines a self-supervised backbone with a supervised 3D detection head. First, a self-supervised scene flow estimation model is trained with cycle consistency. Then, the point cloud encoder of this model is used as the backbone of a single-frame 3D object detection head model. This second 3D object detection model learns to utilize motion representations to distinguish dynamic objects exhibiting different movement patterns. Experiments on KITTI and nuScenes benchmarks show that the proposed self-supervised pre-training increases 3D detection performance significantly. https://github.com/emecercelik/ssl-3d-detection.git
White matter fiber clustering is an important strategy for white matter parcellation, which enables quantitative analysis of brain connections in health and disease. In combination with expert neuroanatomical labeling, data-driven white matter fiber clustering is a powerful tool for creating atlases that can model white matter anatomy across individuals. While widely used fiber clustering approaches have shown good performance using classical unsupervised machine learning techniques, recent advances in deep learning reveal a promising direction toward fast and effective fiber clustering. In this work, we propose a novel deep learning framework for white matter fiber clustering, Deep Fiber Clustering (DFC), which solves the unsupervised clustering problem as a self-supervised learning task with a domain-specific pretext task to predict pairwise fiber distances. This process learns a high-dimensional embedding feature representation for each fiber, regardless of the order of fiber points reconstructed during tractography. We design a novel network architecture that represents input fibers as point clouds and allows the incorporation of additional sources of input information from gray matter parcellation to improve anatomical coherence of clusters. In addition, DFC conducts outlier removal naturally by rejecting fibers with low cluster assignment probability. We evaluate DFC on three independently acquired cohorts, including data from 220 individuals across genders, ages (young and elderly adults), and different health conditions (healthy control and multiple neuropsychiatric disorders). We compare DFC to several state-of-the-art white matter fiber clustering algorithms. Experimental results demonstrate superior performance of DFC in terms of cluster compactness, generalization ability, anatomical coherence, and computational efficiency.