Preventable or undiagnosed visual impairment and blindness affect billion of people worldwide. Automated multi-disease detection models offer great potential to address this problem via clinical decision support in diagnosis. In this work, we proposed an innovative multi-disease detection pipeline for retinal imaging which utilizes ensemble learning to combine the predictive capabilities of several heterogeneous deep convolutional neural network models. Our pipeline includes state-of-the-art strategies like transfer learning, class weighting, real-time image augmentation and Focal loss utilization. Furthermore, we integrated ensemble learning techniques like heterogeneous deep learning models, bagging via 5-fold cross-validation and stacked logistic regression models. Through internal and external evaluation, we were able to validate and demonstrate high accuracy and reliability of our pipeline, as well as the comparability with other state-of-the-art pipelines for retinal disease prediction.
While self-supervised pretraining has proven beneficial for many computer vision tasks, it requires expensive and lengthy computation, large amounts of data, and is sensitive to data augmentation. Prior work demonstrates that models pretrained on datasets dissimilar to their target data, such as chest X-ray models trained on ImageNet, underperform models trained from scratch. Users that lack the resources to pretrain must use existing models with lower performance. This paper explores Hierarchical PreTraining (HPT), which decreases convergence time and improves accuracy by initializing the pretraining process with an existing pretrained model. Through experimentation on 16 diverse vision datasets, we show HPT converges up to 80x faster, improves accuracy across tasks, and improves the robustness of the self-supervised pretraining process to changes in the image augmentation policy or amount of pretraining data. Taken together, HPT provides a simple framework for obtaining better pretrained representations with less computational resources.
Virtual-reality (VR) and augmented-reality (AR) technology is increasingly combined with eye-tracking. This combination broadens both fields and opens up new areas of application, in which visual perception and related cognitive processes can be studied in interactive but still well controlled settings. However, performing a semantic gaze analysis of eye-tracking data from interactive three-dimensional scenes is a resource-intense task, which so far has been an obstacle to economic use. In this paper we present a novel approach which minimizes time and information necessary to annotate volumes of interest (VOIs) by using techniques from object recognition. To do so, we train convolutional neural networks (CNNs) on synthetic data sets derived from virtual models using image augmentation techniques. We evaluate our method in real and virtual environments, showing that the method can compete with state-of-the-art approaches, while not relying on additional markers or preexisting databases but instead offering cross-platform use.
In this work we propose a novel deep-learning approach for age estimation based on face images. We first introduce a dual image augmentation-aggregation approach based on attention. This allows the network to jointly utilize multiple face image augmentations whose embeddings are aggregated by a Transformer-Encoder. The resulting aggregated embedding is shown to better encode the face image attributes. We then propose a probabilistic hierarchical regression framework that combines a discrete probabilistic estimate of age labels, with a corresponding ensemble of regressors. Each regressor is particularly adapted and trained to refine the probabilistic estimate over a range of ages. Our scheme is shown to outperform contemporary schemes and provide a new state-of-the-art age estimation accuracy, when applied to the MORPH II dataset for age estimation. Last, we introduce a bias analysis of state-of-the-art age estimation results.
Data augmentation is an effective technique to improve the generalization of deep neural networks. However, previous data augmentation methods usually treat the augmented samples equally without considering their individual impacts on the model. To address this, for the augmented samples from the same training example, we propose to assign different weights to them. We construct the maximal expected loss which is the supremum over any reweighted loss on augmented samples. Inspired by adversarial training, we minimize this maximal expected loss (MMEL) and obtain a simple and interpretable closed-form solution: more attention should be paid to augmented samples with large loss values (i.e., harder examples). Minimizing this maximal expected loss enables the model to perform well under any reweighting strategy. The proposed method can generally be applied on top of any data augmentation methods. Experiments are conducted on both natural language understanding tasks with token-level data augmentation, and image classification tasks with commonly-used image augmentation techniques like random crop and horizontal flip. Empirical results show that the proposed method improves the generalization performance of the model.
Emerging interests have been brought to recognize previously unseen objects given very few training examples, known as few-shot object detection (FSOD). Recent researches demonstrate that good feature embedding is the key to reach favorable few-shot learning performance. We observe object proposals with different Intersection-of-Union (IoU) scores are analogous to the intra-image augmentation used in contrastive approaches. And we exploit this analogy and incorporate supervised contrastive learning to achieve more robust objects representations in FSOD. We present Few-Shot object detection via Contrastive proposals Encoding (FSCE), a simple yet effective approach to learning contrastive-aware object proposal encodings that facilitate the classification of detected objects. We notice the degradation of average precision (AP) for rare objects mainly comes from misclassifying novel instances as confusable classes. And we ease the misclassification issues by promoting instance level intra-class compactness and inter-class variance via our contrastive proposal encoding loss (CPE loss). Our design outperforms current state-of-the-art works in any shot and all data splits, with up to +8.8% on standard benchmark PASCAL VOC and +2.7% on challenging COCO benchmark. Code is available at: https: //github.com/MegviiDetection/FSCE
Autonomous vehicles rely heavily upon their perception subsystems to see the environment in which they operate. Unfortunately, the effect of variable weather conditions presents a significant challenge to object detection algorithms, and thus it is imperative to test the vehicle extensively in all conditions which it may experience. However, development of robust autonomous vehicle subsystems requires repeatable, controlled testing - while real weather is unpredictable and cannot be scheduled. Real-world testing in adverse conditions is an expensive and time-consuming task, often requiring access to specialist facilities. Simulation is commonly relied upon as a substitute, with increasingly visually realistic representations of the real-world being developed. In the context of the complete autonomous vehicle control pipeline, subsystems downstream of perception need to be tested with accurate recreations of the perception system output, rather than focusing on subjective visual realism of the input - whether in simulation or the real world. This study develops the untapped potential of a lightweight weather augmentation method in an autonomous racing vehicle - focusing not on visual accuracy, but rather the effect upon perception subsystem performance in real time. With minimal adjustment, the prototype developed in this study can replicate the effects of water droplets on the camera lens, and fading light conditions. This approach introduces a latency of less than 8 ms using compute hardware well suited to being carried in the vehicle - rendering it ideal for real-time implementation that can be run during experiments in simulation, and augmented reality testing in the real world.
Deep learning semantic segmentation algorithms can localise abnormalities or opacities from chest radiographs. However, the task of collecting and annotating training data is expensive and requires expertise which remains a bottleneck for algorithm performance. We investigate the effect of image augmentations on reducing the requirement of labelled data in the semantic segmentation of chest X-rays for pneumonia detection. We train fully convolutional network models on subsets of different sizes from the total training data. We apply a different image augmentation while training each model and compare it to the baseline trained on the entire dataset without augmentations. We find that rotate and mixup are the best augmentations amongst rotate, mixup, translate, gamma and horizontal flip, wherein they reduce the labelled data requirement by 70% while performing comparably to the baseline in terms of AUC and mean IoU in our experiments.
Large datasets' availability is catalyzing a rapid expansion of deep learning in general and computer vision in particular. At the same time, in many domains, a sufficient amount of training data is lacking, which may become an obstacle to the practical application of computer vision techniques. This paper challenges small and imbalanced datasets based on the example of a plant phenomics domain. We introduce an image augmentation framework, which enables us to extremely enlarge the number of training samples while providing the data for such tasks as object detection, semantic segmentation, instance segmentation, object counting, image denoising, and classification. We prove that our augmentation method increases model performance when only a few training samples are available. In our experiment, we use the DeepLabV3 model on semantic segmentation tasks with Arabidopsis and Nicotiana tabacum image dataset. The obtained result shows a 9% relative increase in model performance compared to the basic image augmentation techniques.
Robustness is a significant constraint in machine learning models. The performance of the algorithms must not deteriorate when training and testing with slightly different data. Deep neural network models achieve awe-inspiring results in a wide range of applications of computer vision. Still, in the presence of noise or region occlusion, some models exhibit inaccurate performance even with data handled in training. Besides, some experiments suggest deep learning models sometimes use incorrect parts of the input information to perform inference. Activate Image Augmentation (ADA) is an augmentation method that uses interpretability methods to augment the training data and improve its robustness to face the described problems. Although ADA presented interesting results, its original version only used the Vanilla Backpropagation interpretability to train the U-Net model. In this work, we propose an extensive experimental analysis of the interpretability method's impact on ADA. We use five interpretability methods: Vanilla Backpropagation, Guided Backpropagation, GradCam, Guided GradCam, and InputXGradient. The results show that all methods achieve similar performance at the ending of training, but when combining ADA with GradCam, the U-Net model presented an impressive fast convergence.
The shortage of annotated medical images is one of the biggest challenges in the field of medical image computing. Without a sufficient number of training samples, deep learning based models are very likely to suffer from over-fitting problem. The common solution is image manipulation such as image rotation, cropping, or resizing. Those methods can help relieve the over-fitting problem as more training samples are introduced. However, they do not really introduce new images with additional information and may lead to data leakage as the test set may contain similar samples which appear in the training set. To address this challenge, we propose to generate diverse images with generative adversarial network. In this paper, we develop a novel generative method named generative adversarial U-Net , which utilizes both generative adversarial network and U-Net. Different from existing approaches, our newly designed model is domain-free and generalizable to various medical images. Extensive experiments are conducted over eight diverse datasets including computed tomography (CT) scan, pathology, X-ray, etc. The visualization and quantitative results demonstrate the efficacy and good generalization of the proposed method on generating a wide array of high-quality medical images.
Deep neural networks are vulnerable to adversarial examples, which are crafted by adding small, human-imperceptible perturbations to the original images, but make the model output inaccurate predictions. Before deep neural networks are deployed, adversarial attacks can thus be an important method to evaluate and select robust models in safety-critical applications. However, under the challenging black-box setting, the attack success rate, i.e., the transferability of adversarial examples, still needs to be improved. Based on image augmentation methods, we found that random transformation of image brightness can eliminate overfitting in the generation of adversarial examples and improve their transferability. To this end, we propose an adversarial example generation method based on this phenomenon, which can be integrated with Fast Gradient Sign Method (FGSM)-related methods to build a more robust gradient-based attack and generate adversarial examples with better transferability. Extensive experiments on the ImageNet dataset demonstrate the method's effectiveness. Whether on normally or adversarially trained networks, our method has a higher success rate for black-box attacks than other attack methods based on data augmentation. We hope that this method can help to evaluate and improve the robustness of models.
Tuberculosis (TB) is a contagious bacterial airborne disease, and is one of the top 10 causes of death worldwide. According to the World Health Organization (WHO), around 1.8 billion people are infected with TB and 1.6 million deaths were reported in 2018. More importantly,95% of cases and deaths were from developing countries. Yet, TB is a completely curable disease through early diagnosis. To achieve this goal one of the key requirements is efficient utilization of existing diagnostic technologies, among which chest X-ray is the first line of diagnostic tool used for screening for active TB. The presented deep learning pipeline consists of three different state of the art deep learning architectures, to generate, segment and classify lung X-rays. Apart from this image preprocessing, image augmentation, genetic algorithm based hyper parameter tuning and model ensembling were used to to improve the diagnostic process. We were able to achieve classification accuracy of 97.1% (Youden's index-0.941,sensitivity of 97.9% and specificity of 96.2%) which is a considerable improvement compared to the existing work in the literature. In our work, we present an highly accurate, automated TB screening system using chest X-rays, which would be helpful especially for low income countries with low access to qualified medical professionals.
Automatic Target Recognition (ATR) algorithms classify a given Synthetic Aperture Radar (SAR) image into one of the known target classes using a set of training images available for each class. Recently, learning methods have shown to achieve state-of-the-art classification accuracy if abundant training data is available, sampled uniformly over the classes, and their poses. In this paper, we consider the task of ATR with a limited set of training images. We propose a data augmentation approach to incorporate domain knowledge and improve the generalization power of a data-intensive learning algorithm, such as a Convolutional neural network (CNN). The proposed data augmentation method employs a limited persistence sparse modeling approach, capitalizing on commonly observed characteristics of wide-angle synthetic aperture radar (SAR) imagery. Specifically, we exploit the sparsity of the scattering centers in the spatial domain and the smoothly-varying structure of the scattering coefficients in the azimuthal domain to solve the ill-posed problem of over-parametrized model fitting. Using this estimated model, we synthesize new images at poses and sub-pixel translations not available in the given data to augment CNN's training data. The experimental results show that for the training data starved region, the proposed method provides a significant gain in the resulting ATR algorithm's generalization performance.
Building instance segmentation models that are data-efficient and can handle rare object categories is an important challenge in computer vision. Leveraging data augmentations is a promising direction towards addressing this challenge. Here, we perform a systematic study of the Copy-Paste augmentation ([13, 12]) for instance segmentation where we randomly paste objects onto an image. Prior studies on Copy-Paste relied on modeling the surrounding visual context for pasting the objects. However, we find that the simple mechanism of pasting objects randomly is good enough and can provide solid gains on top of strong baselines. Furthermore, we show Copy-Paste is additive with semi-supervised methods that leverage extra data through pseudo labeling (e.g. self-training). On COCO instance segmentation, we achieve 49.1 mask AP and 57.3 box AP, an improvement of +0.6 mask AP and +1.5 box AP over the previous state-of-the-art. We further demonstrate that Copy-Paste can lead to significant improvements on the LVIS benchmark. Our baseline model outperforms the LVIS 2020 Challenge winning entry by +3.6 mask AP on rare categories.
The purpose of this paper is to design a solution to the problem of facial recognition by use of convolutional neural networks, with the intention of applying the solution in a camera-based home-entry access control system. More specifically, the paper focuses on solving the supervised classification problem of taking images of people as input and classifying the person in the image as one of the authors or not. Two approaches are proposed: (1) building and training a neural network called WoodNet from scratch and (2) leveraging transfer learning by utilizing a network pre-trained on the ImageNet database and adapting it to this project's data and classes. In order to train the models to recognize the authors, a dataset containing more than 150 000 images has been created, balanced over the authors and others. Image extraction from videos and image augmentation techniques were instrumental for dataset creation. The results are two models classifying the individuals in the dataset with high accuracy, achieving over 99% accuracy on held-out test data. The pre-trained model fitted significantly faster than WoodNet, and seems to generalize better. However, these results come with a few caveats. Because of the way the dataset was compiled, as well as the high accuracy, one has reason to believe the models over-fitted to the data to some degree. An added consequence of the data compilation method is that the test dataset may not be sufficiently different from the training data, limiting its ability to validate generalization of the models. However, utilizing the models in a web-cam based system, classifying faces in real-time, shows promising results and indicates that the models generalized fairly well for at least some of the classes (see the accompanying video).
Pavement condition evaluation is essential to time the preventative or rehabilitative actions and control distress propagation. Failing to conduct timely evaluations can lead to severe structural and financial loss of the infrastructure and complete reconstructions. Automated computer-aided surveying measures can provide a database of road damage patterns and their locations. This database can be utilized for timely road repairs to gain the minimum cost of maintenance and the asphalt's maximum durability. This paper introduces a deep learning-based surveying scheme to analyze the image-based distress data in real-time. A database consisting of a diverse population of crack distress types such as longitudinal, transverse, and alligator cracks, photographed using mobile-device is used. Then, a family of efficient and scalable models that are tuned for pavement crack detection is trained, and various augmentation policies are explored. Proposed models, resulted in F1-scores, ranging from 52% to 56%, and average inference time from 178-10 images per second. Finally, the performance of the object detectors are examined, and error analysis is reported against various images. The source code is available at https://github.com/mahdi65/roadDamageDetection2020.
In this paper we present a review of the Kornia differentiable data augmentation (DDA) module for both for spatial (2D) and volumetric (3D) tensors. This module leverages differentiable computer vision solutions from Kornia, with an aim of integrating data augmentation (DA) pipelines and strategies to existing PyTorch components (e.g. autograd for differentiability, optim for optimization). In addition, we provide a benchmark comparing different DA frameworks and a short review for a number of approaches that make use of Kornia DDA.
Gastrointestinal (GI) pathologies are periodically screened, biopsied, and resected using surgical tools. Usually the procedures and the treated or resected areas are not specifically tracked or analysed during or after colonoscopies. Information regarding disease borders, development and amount and size of the resected area get lost. This can lead to poor follow-up and bothersome reassessment difficulties post-treatment. To improve the current standard and also to foster more research on the topic we have released the ``Kvasir-Instrument'' dataset which consists of $590$ annotated frames containing GI procedure tools such as snares, balloons and biopsy forceps, etc. Beside of the images, the dataset includes ground truth masks and bounding boxes and has been verified by two expert GI endoscopists. Additionally, we provide a baseline for the segmentation of the GI tools to promote research and algorithm development. We obtained a dice coefficient score of 0.9158 and a Jaccard index of 0.8578 using a classical U-Net architecture. A similar dice coefficient score was observed for DoubleUNet. The qualitative results showed that the model did not work for the images with specularity and the frames with multiple instruments, while the best result for both methods was observed on all other types of images. Both, qualitative and quantitative results show that the model performs reasonably good, but there is a large potential for further improvements. Benchmarking using the dataset provides an opportunity for researchers to contribute to the field of automatic endoscopic diagnostic and therapeutic tool segmentation for GI endoscopy.
Over the last several years, research on facial recognition based on Deep Neural Network has evolved with approaches like task-specific loss functions, image normalization and augmentation, network architectures, etc. However, there have been few approaches with attention to how human faces differ from person to person. Premising that inter-personal differences are found both generally and locally on the human face, I propose FusiformNet, a novel framework for feature extraction that leverages the nature of discriminative facial features. Tested on Image-Unrestricted setting of Labeled Faces in the Wild benchmark, this method achieved a state-of-the-art accuracy of 96.67% without labeled outside data, image augmentation, normalization, or special loss functions. Likewise, the method also performed on a par with previous state-of-the-arts when pre-trained on CASIA-WebFace dataset. Considering its ability to extract both general and local facial features, the utility of FusiformNet may not be limited to facial recognition but also extend to other DNN-based tasks.