CNN-based Real-time Dense Face Reconstruction with Inverse-rendered Photo-realistic Face Images

With the powerfulness of convolution neural networks (CNN), CNN based face reconstruction has recently shown promising performance in reconstructing detailed face shape from 2D face images. The success of CNN-based methods relies on a large number of labeled data. The state-of-the-art synthesizes such data using a coarse morphable face model, which however has difficulty to generate detailed photo-realistic images of faces (with wrinkles). This paper presents a novel face data generation method. Specifically, we render a large number of photo-realistic face images with different attributes based on inverse rendering. Furthermore, we construct a fine-detailed face image dataset by transferring different scales of details from one image to another. We also construct a large number of video-type adjacent frame pairs by simulating the distribution of real video data. With these nicely constructed datasets, we propose a coarse-to-fine learning framework consisting of three convolutional networks. The networks are trained for real-time detailed 3D face reconstruction from monocular video as well as from a single image. Extensive experimental results demonstrate that our framework can produce high-quality reconstruction but with much less computation time compared to the state-of-the-art. Moreover, our method is robust to pose, expression and lighting due to the diversity of data.

Piece-wise Matching Layer in Representation Learning for ECG Classification

This paper proposes piece-wise matching layer as a novel layer in representation learning methods for electrocardiogram (ECG) classification. Despite the remarkable performance of representation learning methods in the analysis of time series, there are still several challenges associated with these methods ranging from the complex structures of methods, the lack of generality of solutions, the need for expert knowledge, and large-scale training datasets. We introduce the piece-wise matching layer that works based on two levels to address some of the aforementioned challenges. At the first level, a set of morphological, statistical, and frequency features and comparative forms of them are computed based on each periodic part and its neighbors. At the second level, these features are modified by predefined transformation functions based on a receptive field scenario. Several scenarios of offline processing, incremental processing, fixed sliding receptive field, and event-based triggering receptive field can be implemented based on the choice of length and mechanism of indicating the receptive field. We propose dynamic time wrapping as a mechanism that indicates a receptive field based on event triggering tactics. To evaluate the performance of this method in time series analysis, we applied the proposed layer in two publicly available datasets of PhysioNet competitions in 2015 and 2017 where the input data is ECG signal. We compared the performance of our method against a variety of known tuned methods from expert knowledge, machine learning, deep learning methods, and the combination of them. The proposed approach improves the state of the art in two known completions 2015 and 2017 around 4% and 7% correspondingly while it does not rely on in advance knowledge of the classes or the possible places of arrhythmia.

Boost AI Power: Data Augmentation Strategies with unlabelled Data and Conformal Prediction, a Case in Alternative Herbal Medicine Discrimination with Electronic Nose

Electronic nose proves its effectiveness in alternativeherbal medicine classification, but due to the supervised learn-ing nature, previous research relies on the labelled training data,which are time-costly and labor-intensive to collect. Consideringthe training data inadequacy in real-world applications, this studyaims to improve classification accuracy via data augmentationstrategies. We stimulated two scenarios to investigate the effective-ness of five data augmentation strategies under different trainingdata inadequacy: in the noise-free scenario, different availability ofunlabelled data were simulated, and in the noisy scenario, differentlevels of Gaussian noises and translational shifts were added tosimulate sensor drifts. The augmentation strategies: noise-addingdata augmentation, semi-supervised learning, classifier-based online learning, inductive conformal prediction (ICP) onlinelearning and the novel ensemble ICP online learning proposed in this study, were compared against supervised learningbaseline, with Linear Discriminant Analysis (LDA) and Support Vector Machine (SVM) as the classifiers. We found thatat least one strategies significantly improved the classification accuracy with LDA(p<=0.05) and showed non-decreasingclassification accuracy with SVM in each tasks. Moreover, our novel strategy: ensemble ICP online learning outperformedthe others by showing non-decreasing classification accuracy on all tasks and significant improvement on most tasks(25/36 tasks,p<=0.05). This study provides a systematic analysis over augmentation strategies, and we provided userswith recommended strategies under specific circumstances. Furthermore, our newly proposed strategy showed botheffectiveness and robustness in boosting the classification model generalizability, which can also be further employed inother machine learning applications.

Coordinated Aerial-Ground Robot Exploration via Monte-Carlo View Quality Rendering

We present a framework for a ground-aerial robotic team to explore large, unstructured, and unknown environments. In such exploration problems, the effectiveness of existing exploration-boosting heuristics often scales poorly with the environments' size and complexity. This work proposes a novel framework combining incremental frontier distribution, goal selection with Monte-Carlo view quality rendering, and an automatic-differentiable information gain measure to improve exploration efficiency. Simulated with multiple complex environments, we demonstrate that the proposed method effectively utilizes collaborative aerial and ground robots, consistently guides agents to informative viewpoints, improves exploration paths' information gain, and reduces planning time.

Analysis of Drifting Features

The notion of concept drift refers to the phenomenon that the distribution, which is underlying the observed data, changes over time. We are interested in an identification of those features, that are most relevant for the observed drift. We distinguish between drift inducing features, for which the observed feature drift cannot be explained by any other feature, and faithfully drifting features, which correlate with the present drift of other features. This notion gives rise to minimal subsets of the feature space, which are able to characterize the observed drift as a whole. We relate this problem to the problems of feature selection and feature relevance learning, which allows us to derive a detection algorithm. We demonstrate its usefulness on different benchmarks.

Benchmarking of eight recurrent neural network variants for breath phase and adventitious sound detection on a self-developed open-access lung sound database-HF_Lung_V1

A reliable, remote, and continuous real-time respiratory sound monitor with automated respiratory sound analysis ability is urgently required in many clinical scenarios-such as in monitoring disease progression of coronavirus disease 2019-to replace conventional auscultation with a handheld stethoscope. However, a robust computerized respiratory sound analysis algorithm has not yet been validated in practical applications. In this study, we developed a lung sound database (HF_Lung_V1) comprising 9,765 audio files of lung sounds (duration of 15 s each), 34,095 inhalation labels, 18,349 exhalation labels, 13,883 continuous adventitious sound (CAS) labels (comprising 8,457 wheeze labels, 686 stridor labels, and 4,740 rhonchi labels), and 15,606 discontinuous adventitious sound labels (all crackles). We conducted benchmark tests for long short-term memory (LSTM), gated recurrent unit (GRU), bidirectional LSTM (BiLSTM), bidirectional GRU (BiGRU), convolutional neural network (CNN)-LSTM, CNN-GRU, CNN-BiLSTM, and CNN-BiGRU models for breath phase detection and adventitious sound detection. We also conducted a performance comparison between the LSTM-based and GRU-based models, between unidirectional and bidirectional models, and between models with and without a CNN. The results revealed that these models exhibited adequate performance in lung sound analysis. The GRU-based models outperformed, in terms of F1 scores and areas under the receiver operating characteristic curves, the LSTM-based models in most of the defined tasks. Furthermore, all bidirectional models outperformed their unidirectional counterparts. Finally, the addition of a CNN improved the accuracy of lung sound analysis, especially in the CAS detection tasks.

Disease control as an optimization problem

Traditionally, expert epidemiologists devise policies for disease control through a mixture of intuition and brute force. Namely, they use their know-how to narrow down the set of logically conceivable policies to a small family described by a few parameters, following which they conduct a grid search to identify the optimal policy within the set. This scheme is not scalable, in the sense that, when used to optimize over policies which depend on many parameters, it would likely fail to output an optimal disease policy in time for its implementation. In this article, we use techniques from convex optimization theory and machine learning to conduct optimizations over disease policies described by hundreds of parameters. We illustrate our approach by minimizing the total time required to eradicate COVID-19 within the Susceptible-Exposed-Infected-Recovered (SEIR) model proposed by Kissler et al. (March, 2020). We observe that optimal policies are typically very complex, and thus unlikely to be discovered by a human agent.

Deep Image Prior for Sparse-sampling Photoacoustic Microscopy

Photoacoustic microscopy (PAM) is an emerging method for imaging both structural and functional information without the need for exogenous contrast agents. However, state-of-the-art PAM faces a tradeoff between imaging speed and spatial sampling density within the same field-of-view (FOV). Limited by the pulsed laser's repetition rate, the imaging speed is inversely proportional to the total number of effective pixels. To cover the same FOV in a shorter amount of time with the same PAM hardware, there is currently no other option than to decrease spatial sampling density (i.e., sparse sampling). Deep learning methods have recently been used to improve sparsely sampled PAM images; however, these methods often require time-consuming pre-training and a large training dataset that has fully sampled, co-registered ground truth. In this paper, we propose using a method known as "deep image prior" to improve the image quality of sparsely sampled PAM images. The network does not need prior learning or fully sampled ground truth, making its implementation more flexible and much quicker. Our results show promising improvement in PA vasculature images with as few as 2% of the effective pixels. Our deep image prior approach produces results that outperform interpolation methods and can be readily translated to other high-speed, sparse-sampling imaging modalities.

An Active Learning Method for Diabetic Retinopathy Classification with Uncertainty Quantification

In recent years, deep learning (DL) techniques have provided state-of-the-art performance on different medical imaging tasks. However, the availability of good quality annotated medical data is very challenging due to involved time constraints and the availability of expert annotators, e.g., radiologists. In addition, DL is data-hungry and their training requires extensive computational resources. Another problem with DL is their black-box nature and lack of transparency on its inner working which inhibits causal understanding and reasoning. In this paper, we jointly address these challenges by proposing a hybrid model, which uses a Bayesian convolutional neural network (BCNN) for uncertainty quantification, and an active learning approach for annotating the unlabelled data. The BCNN is used as a feature descriptor and these features are then used for training a model, in an active learning setting. We evaluate the proposed framework for diabetic retinopathy classification problem and have achieved state-of-the-art performance in terms of different metrics.

HiPPO: Recurrent Memory with Optimal Polynomial Projections

A central problem in learning from sequential data is representing cumulative history in an incremental fashion as more data is processed. We introduce a general framework (HiPPO) for the online compression of continuous signals and discrete time series by projection onto polynomial bases. Given a measure that specifies the importance of each time step in the past, HiPPO produces an optimal solution to a natural online function approximation problem. As special cases, our framework yields a short derivation of the recent Legendre Memory Unit (LMU) from first principles, and generalizes the ubiquitous gating mechanism of recurrent neural networks such as GRUs. This formal framework yields a new memory update mechanism (HiPPO-LegS) that scales through time to remember all history, avoiding priors on the timescale. HiPPO-LegS enjoys the theoretical benefits of timescale robustness, fast updates, and bounded gradients. By incorporating the memory dynamics into recurrent neural networks, HiPPO RNNs can empirically capture complex temporal dependencies. On the benchmark permuted MNIST dataset, HiPPO-LegS sets a new state-of-the-art accuracy of 98.3%. Finally, on a novel trajectory classification task testing robustness to out-of-distribution timescales and missing data, HiPPO-LegS outperforms RNN and neural ODE baselines by 25-40% accuracy.