Despite the recent progress in image dehazing, several problems remain largely unsolved such as robustness for varying scenes, the visual quality of reconstructed images, and effectiveness and flexibility for applications. To tackle these problems, we propose a new deep network architecture for single image dehazing called DR-Net. Our model consists of three main subnetworks: a transmission prediction network that predicts transmission map for the input image, a haze removal network that reconstructs latent image steered by the transmission map, and a refinement network that enhances the details and color properties of the dehazed result via weakly supervised learning. Compared to previous methods, our method advances in three aspects: (i) pure data-driven model; (ii) the end-to-end system; (iii) superior robustness, accuracy, and applicability. Extensive experiments demonstrate that our DR-Net outperforms the state-of-the-art methods on both synthetic and real images in qualitative and quantitative metrics. Additionally, the utility of DR-Net has been illustrated by its potential usage in several important computer vision tasks.
Food classification from images is a fine-grained classification problem. Manual curation of food images is cost, time and scalability prohibitive. On the other hand, web data is available freely but contains noise. In this paper, we address the problem of classifying food images with minimal data curation. We also tackle a key problems with food images from the web where they often have multiple cooccuring food types but are weakly labeled with a single label. We first demonstrate that by sequentially adding a few manually curated samples to a larger uncurated dataset from two web sources, the top-1 classification accuracy increases from 50.3% to 72.8%. To tackle the issue of weak labels, we augment the deep model with Weakly Supervised learning (WSL) that results in an increase in performance to 76.2%. Finally, we show some qualitative results to provide insights into the performance improvements using the proposed ideas.
Spatially aligning medical images from different modalities remains a challenging task, especially for intraoperative applications that require fast and robust algorithms. We propose a weakly-supervised, label-driven formulation for learning 3D voxel correspondence from higher-level label correspondence, thereby bypassing classical intensity-based image similarity measures. During training, a convolutional neural network is optimised by outputting a dense displacement field (DDF) that warps a set of available anatomical labels from the moving image to match their corresponding counterparts in the fixed image. These label pairs, including solid organs, ducts, vessels, point landmarks and other ad hoc structures, are only required at training time and can be spatially aligned by minimising a cross-entropy function of the warped moving label and the fixed label. During inference, the trained network takes a new image pair to predict an optimal DDF, resulting in a fully-automatic, label-free, real-time and deformable registration. For interventional applications where large global transformation prevails, we also propose a neural network architecture to jointly optimise the global- and local displacements. Experiment results are presented based on cross-validating registrations of 111 pairs of T2-weighted magnetic resonance images and 3D transrectal ultrasound images from prostate cancer patients with a total of over 4000 anatomical labels, yielding a median target registration error of 4.2 mm on landmark centroids and a median Dice of 0.88 on prostate glands.
Image segmentation is a fundamental problem in medical image analysis. In recent years, deep neural networks achieve impressive performances on many medical image segmentation tasks by supervised learning on large manually annotated data. However, expert annotations on big medical datasets are tedious, expensive or sometimes unavailable. Weakly supervised learning could reduce the effort for annotation but still required certain amounts of expertise. Recently, deep learning shows a potential to produce more accurate predictions than the original erroneous labels. Inspired by this, we introduce a very weakly supervised learning method, for cystic lesion detection and segmentation in lung CT images, without any manual annotation. Our method works in a self-learning manner, where segmentation generated in previous steps (first by unsupervised segmentation then by neural networks) is used as ground truth for the next level of network learning. Experiments on a cystic lung lesion dataset show that the deep learning could perform better than the initial unsupervised annotation, and progressively improve itself after self-learning.
This paper investigates a fundamental problem of scene understanding: how to parse a scene image into a structured configuration (i.e., a semantic object hierarchy with object interaction relations). We propose a deep architecture consisting of two networks: i) a convolutional neural network (CNN) extracting the image representation for pixel-wise object labeling and ii) a recursive neural network (RsNN) discovering the hierarchical object structure and the inter-object relations. Rather than relying on elaborative annotations (e.g., manually labeled semantic maps and relations), we train our deep model in a weakly-supervised learning manner by leveraging the descriptive sentences of the training images. Specifically, we decompose each sentence into a semantic tree consisting of nouns and verb phrases, and apply these tree structures to discover the configurations of the training images. Once these scene configurations are determined, then the parameters of both the CNN and RsNN are updated accordingly by back propagation. The entire model training is accomplished through an Expectation-Maximization method. Extensive experiments show that our model is capable of producing meaningful scene configurations and achieving more favorable scene labeling results on two benchmarks (i.e., PASCAL VOC 2012 and SYSU-Scenes) compared with other state-of-the-art weakly-supervised deep learning methods. In particular, SYSU-Scenes contains more than 5000 scene images with their semantic sentence descriptions, which is created by us for advancing research on scene parsing.
A persistent challenge in practical classification tasks is that labeled training sets are not always available. In particle physics, this challenge is surmounted by the use of simulations. These simulations accurately reproduce most features of data, but cannot be trusted to capture all of the complex correlations exploitable by modern machine learning methods. Recent work in weakly supervised learning has shown that simple, low-dimensional classifiers can be trained using only the impure mixtures present in data. Here, we demonstrate that complex, high-dimensional classifiers can also be trained on impure mixtures using weak supervision techniques, with performance comparable to what could be achieved with pure samples. Using weak supervision will therefore allow us to avoid relying exclusively on simulations for high-dimensional classification. This work opens the door to a new regime whereby complex models are trained directly on data, providing direct access to probe the underlying physics.
Analysis of histopathology slides is a critical step for many diagnoses, and in particular in oncology where it defines the gold standard. In the case of digital histopathological analysis, highly trained pathologists must review vast whole-slide-images of extreme digital resolution ($100,000^2$ pixels) across multiple zoom levels in order to locate abnormal regions of cells, or in some cases single cells, out of millions. The application of deep learning to this problem is hampered not only by small sample sizes, as typical datasets contain only a few hundred samples, but also by the generation of ground-truth localized annotations for training interpretable classification and segmentation models. We propose a method for disease localization in the context of weakly supervised learning, where only image-level labels are available during training. Even without pixel-level annotations, we are able to demonstrate performance comparable with models trained with strong annotations on the Camelyon-16 lymph node metastases detection challenge. We accomplish this through the use of pre-trained deep convolutional networks, feature embedding, as well as learning via top instances and negative evidence, a multiple instance learning technique from the field of semantic segmentation and object detection.
Patient pain can be detected highly reliably from facial expressions using a set of facial muscle-based action units (AUs) defined by the Facial Action Coding System (FACS). A key characteristic of facial expression of pain is the simultaneous occurrence of pain-related AU combinations, whose automated detection would be highly beneficial for efficient and practical pain monitoring. Existing general Automated Facial Expression Recognition (AFER) systems prove inadequate when applied specifically for detecting pain as they either focus on detecting individual pain-related AUs but not on combinations or they seek to bypass AU detection by training a binary pain classifier directly on pain intensity data but are limited by lack of enough labeled data for satisfactory training. In this paper, we propose a new approach that mimics the strategy of human coders of decoupling pain detection into two consecutive tasks: one performed at the individual video-frame level and the other at video-sequence level. Using state-of-the-art AFER tools to detect single AUs at the frame level, we propose two novel data structures to encode AU combinations from single AU scores. Two weakly supervised learning frameworks namely multiple instance learning (MIL) and multiple clustered instance learning (MCIL) are employed corresponding to each data structure to learn pain from video sequences. Experimental results show an 87% pain recognition accuracy with 0.94 AUC (Area Under Curve) on the UNBC-McMaster Shoulder Pain Expression dataset. Tests on long videos in a lung cancer patient video dataset demonstrates the potential value of the proposed system for pain monitoring in clinical settings.
Supervised object detection and semantic segmentation require object or even pixel level annotations. When there exist image level labels only, it is challenging for weakly supervised algorithms to achieve accurate predictions. The accuracy achieved by top weakly supervised algorithms is still significantly lower than their fully supervised counterparts. In this paper, we propose a novel weakly supervised curriculum learning pipeline for multi-label object recognition, detection and semantic segmentation. In this pipeline, we first obtain intermediate object localization and pixel labeling results for the training images, and then use such results to train task-specific deep networks in a fully supervised manner. The entire process consists of four stages, including object localization in the training images, filtering and fusing object instances, pixel labeling for the training images, and task-specific network training. To obtain clean object instances in the training images, we propose a novel algorithm for filtering, fusing and classifying object instances collected from multiple solution mechanisms. In this algorithm, we incorporate both metric learning and density-based clustering to filter detected object instances. Experiments show that our weakly supervised pipeline achieves state-of-the-art results in multi-label image classification as well as weakly supervised object detection and very competitive results in weakly supervised semantic segmentation on MS-COCO, PASCAL VOC 2007 and PASCAL VOC 2012.
Weakly supervised learning with only coarse labels can obtain visual explanations of deep neural network such as attention maps by back-propagating gradients. These attention maps are then available as priors for tasks such as object localization and semantic segmentation. In one common framework we address three shortcomings of previous approaches in modeling such attention maps: We (1) first time make attention maps an explicit and natural component of the end-to-end training, (2) provide self-guidance directly on these maps by exploring supervision form the network itself to improve them, and (3) seamlessly bridge the gap between using weak and extra supervision if available. Despite its simplicity, experiments on the semantic segmentation task demonstrate the effectiveness of our methods. We clearly surpass the state-of-the-art on Pascal VOC 2012 val. and test set. Besides, the proposed framework provides a way not only explaining the focus of the learner but also feeding back with direct guidance towards specific tasks. Under mild assumptions our method can also be understood as a plug-in to existing weakly supervised learners to improve their generalization performance.
An intuition on human segmentation is that when a human is moving in a video, the video-context (e.g., appearance and motion clues) may potentially infer reasonable mask information for the whole human body. Inspired by this, based on popular deep convolutional neural networks (CNN), we explore a very-weakly supervised learning framework for human segmentation task, where only an imperfect human detector is available along with massive weakly-labeled YouTube videos. In our solution, the video-context guided human mask inference and CNN based segmentation network learning iterate to mutually enhance each other until no further improvement gains. In the first step, each video is decomposed into supervoxels by the unsupervised video segmentation. The superpixels within the supervoxels are then classified as human or non-human by graph optimization with unary energies from the imperfect human detection results and the predicted confidence maps by the CNN trained in the previous iteration. In the second step, the video-context derived human masks are used as direct labels to train CNN. Extensive experiments on the challenging PASCAL VOC 2012 semantic segmentation benchmark demonstrate that the proposed framework has already achieved superior results than all previous weakly-supervised methods with object class or bounding box annotations. In addition, by augmenting with the annotated masks from PASCAL VOC 2012, our method reaches a new state-of-the-art performance on the human segmentation task.
We propose a Multi-Instance-Learning (MIL) approach for weakly-supervised learning problems, where a training set is formed by bags (sets of feature vectors or instances) and only labels at bag-level are provided. Specifically, we consider the Multi-Instance Dynamic-Ordinal-Regression (MI-DOR) setting, where the instance labels are naturally represented as ordinal variables and bags are structured as temporal sequences. To this end, we propose Multi-Instance Dynamic Ordinal Random Fields (MI-DORF). In this framework, we treat instance-labels as temporally-dependent latent variables in an Undirected Graphical Model. Different MIL assumptions are modelled via newly introduced high-order potentials relating bag and instance-labels within the energy function of the model. We also extend our framework to address the Partially-Observed MI-DOR problems, where a subset of instance labels are available during training. We show on the tasks of weakly-supervised facial behavior analysis, Facial Action Unit (DISFA dataset) and Pain (UNBC dataset) Intensity estimation, that the proposed framework outperforms alternative learning approaches. Furthermore, we show that MIDORF can be employed to reduce the data annotation efforts in this context by large-scale.
In this paper, we introduce a new dataset for student engagement detection and localization. Digital revolution has transformed the traditional teaching procedure and a result analysis of the student engagement in an e-learning environment would facilitate effective task accomplishment and learning. Well known social cues of engagement/disengagement can be inferred from facial expressions, body movements and gaze pattern. In this paper, student's response to various stimuli videos are recorded and important cues are extracted to estimate variations in engagement level. In this paper, we study the association of a subject's behavioral cues with his/her engagement level, as annotated by labelers. We then localize engaging/non-engaging parts in the stimuli videos using a deep multiple instance learning based framework, which can give useful insight into designing Massive Open Online Courses (MOOCs) video material. Recognizing the lack of any publicly available dataset in the domain of user engagement, a new `in the wild' dataset is created to study the subject engagement problem. The dataset contains 195 videos captured from 78 subjects which is about 16.5 hours of recording. We present detailed baseline results using different classifiers ranging from traditional machine learning to deep learning based approaches. The subject independent analysis is performed so that it can be generalized to new users. The problem of engagement prediction is modeled as a weakly supervised learning problem. The dataset is manually annotated by different labelers for four levels of engagement independently and the correlation studies between annotated and predicted labels of videos by different classifiers is reported. This dataset creation is an effort to facilitate research in various e-learning environments such as intelligent tutoring systems, MOOCs, and others.
Confocal Laser Endomicroscope (CLE) is a novel handheld fluorescence imaging device that has shown promise for rapid intraoperative diagnosis of brain tumor tissue. Currently CLE is capable of image display only and lacks an automatic system to aid the surgeon in analyzing the images. The goal of this project was to develop a computer-aided diagnostic approach for CLE imaging of human glioma with feature localization function. Despite the tremendous progress in object detection and image segmentation methods in recent years, most of such methods require large annotated datasets for training. However, manual annotation of thousands of histopathological images by physicians is costly and time consuming. To overcome this problem, we propose a Weakly-Supervised Learning (WSL)-based model for feature localization that trains on image-level annotations, and then localizes incidences of a class-of-interest in the test image. We developed a novel convolutional neural network for diagnostic features localization from CLE images by employing a novel multiscale activation map that is laterally inhibited and collaterally integrated. To validate our method, we compared proposed model's output to the manual annotation performed by four neurosurgeons on test images. Proposed model achieved 88% mean accuracy and 86% mean intersection over union on intermediate features and 87% mean accuracy and 88% mean intersection over union on restrictive fine features, while outperforming other state of the art methods tested. This system can improve accuracy and efficiency in characterization of CLE images of glioma tissue during surgery, augment intraoperative decision-making process regarding tumor margin and affect resection rates.
Audio content analysis in terms of sound events is an important research problem for a variety of applications. Recently, the development of weak labeling approaches for audio or sound event detection (AED) and availability of large scale weakly labeled dataset have finally opened up the possibility of large scale AED. However, a deeper understanding of how weak labels affect the learning for sound events is still missing from literature. In this work, we first describe a CNN based approach for weakly supervised training of audio events. The approach follows some basic design principle desirable in a learning method relying on weakly labeled audio. We then describe important characteristics, which naturally arise in weakly supervised learning of sound events. We show how these aspects of weak labels affect the generalization of models. More specifically, we study how characteristics such as label density and corruption of labels affects weakly supervised training for audio events. We also study the feasibility of directly obtaining weak labeled data from the web without any manual label and compare it with a dataset which has been manually labeled. The analysis and understanding of these factors should be taken into picture in the development of future weak label learning methods. Audioset, a large scale weakly labeled dataset for sound events is used in our experiments.
Weakly-supervised learning based on, e.g., partially labelled images or image-tags, is currently attracting significant attention in CNN segmentation as it can mitigate the need for full and laborious pixel/voxel annotations. Enforcing high-order (global) inequality constraints on the network output (for instance, to constrain the size of the target region) can leverage unlabeled data, guiding the training process with domain-specific knowledge. Inequality constraints are very flexible because they do not assume exact prior knowledge. However, constrained Lagrangian dual optimization has been largely avoided in deep networks, mainly for computational tractability reasons. To the best of our knowledge, the method of [Pathak et al., 2015] is the only prior work that addresses deep CNNs with linear constraints in weakly supervised segmentation. It uses the constraints to synthesize fully-labeled training masks (proposals) from weak labels, mimicking full supervision and facilitating dual optimization. We propose to introduce a differentiable penalty, which enforces inequality constraints directly in the loss function, avoiding expensive Lagrangian dual iterates and proposal generation. From constrained-optimization perspective, our simple penalty-based approach is not optimal as there is no guarantee that the constraints are satisfied. However, surprisingly, it yields substantially better results than the Lagrangian-based constrained CNNs in [Pathak et al., 2015], while reducing the computational demand for training. By annotating only a small fraction of the pixels, the proposed approach can reach a level of segmentation performance that is comparable to full supervision on three separate tasks. While our experiments focused on basic linear constraints such as the target-region size and image tags, our framework can be easily extended to other non-linear constraints.
We address the problem of 3D shape completion from sparse and noisy point clouds, a fundamental problem in computer vision and robotics. Recent approaches are either data-driven or learning-based: Data-driven approaches rely on a shape model whose parameters are optimized to fit the observations; Learning-based approaches, in contrast, avoid the expensive optimization step by learning to directly predict complete shapes from incomplete observations in a fully-supervised setting. However, full supervision is often not available in practice. In this work, we propose a weakly-supervised learning-based approach to 3D shape completion which neither requires slow optimization nor direct supervision. While we also learn a shape prior on synthetic data, we amortize, i.e., learn, maximum likelihood fitting using deep neural networks resulting in efficient shape completion without sacrificing accuracy. On synthetic benchmarks based on ShapeNet and ModelNet as well as on real robotics data from KITTI and Kinect, we demonstrate that the proposed amortized maximum likelihood approach is able to compete with recent fully supervised baselines and outperforms data-driven approaches, while requiring less supervision and being significantly faster.
In this paper, we focus on weakly supervised learning with noisy training data for both classification and regression problems.We assume that the training outputs are collected from a mixture of a target and correlated noise distributions.Our proposed method simultaneously estimates the target distribution and the quality of each data which is defined as the correlation between the target and data generating distributions.The cornerstone of the proposed method is a Cholesky Block that enables modeling dependencies among mixture distributions in a differentiable manner where we maintain the distribution over the network weights.We first provide illustrative examples in both regression and classification tasks to show the effectiveness of the proposed method.Then, the proposed method is extensively evaluated in a number of experiments where we show that it constantly shows comparable or superior performances compared to existing baseline methods in the handling of noisy data.
Multiple modalities often co-occur when describing natural phenomena. Learning a joint representation of these modalities should yield deeper and more useful representations. Previous generative approaches to multi-modal input either do not learn a joint distribution or require additional computation to handle missing data. Here, we introduce a multimodal variational autoencoder (MVAE) that uses a product-of-experts inference network and a sub-sampled training paradigm to solve the multi-modal inference problem. Notably, our model shares parameters to efficiently learn under any combination of missing modalities. We apply the MVAE on four datasets and match state-of-the-art performance using many fewer parameters. In addition, we show that the MVAE is directly applicable to weakly-supervised learning, and is robust to incomplete supervision. We then consider two case studies, one of learning image transformations---edge detection, colorization, segmentation---as a set of modalities, followed by one of machine translation between two languages. We find appealing results across this range of tasks.
We consider the task of training classifiers without labels. We propose a weakly supervised method---adversarial label learning---that trains classifiers to perform well against an adversary that chooses labels for training data. The weak supervision constrains what labels the adversary can choose. The method therefore minimizes an upper bound of the classifier's error rate using projected primal-dual subgradient descent. Minimizing this bound protects against bias and dependencies in the weak supervision. Experiments on three real datasets show that our method can train without labels and outperforms other approaches for weakly supervised learning.