A novel unsupervised deep learning method is developed to identify individual-specific large scale brain functional networks (FNs) from resting-state fMRI (rsfMRI) in an end-to-end learning fashion. Our method leverages deep Encoder-Decoder networks and conventional brain decomposition models to identify individual-specific FNs in an unsupervised learning framework and facilitate fast inference for new individuals with one forward pass of the deep network. Particularly, convolutional neural networks (CNNs) with an Encoder-Decoder architecture are adopted to identify individual-specific FNs from rsfMRI data by optimizing their data fitting and sparsity regularization terms that are commonly used in brain decomposition models. Moreover, a time-invariant representation learning module is designed to learn features invariant to temporal orders of time points of rsfMRI data. The proposed method has been validated based on a large rsfMRI dataset and experimental results have demonstrated that our method could obtain individual-specific FNs which are consistent with well-established FNs and are informative for predicting brain age, indicating that the individual-specific FNs identified truly captured the underlying variability of individualized functional neuroanatomy.
We present a feature-free photogrammetric technique that enables quantitative 3D mesoscopic (mm-scale height variation) imaging with tens-of-micron accuracy from sequences of images acquired by a smartphone at close range (several cm) under freehand motion without additional hardware. Our end-to-end, pixel-intensity-based approach jointly registers and stitches all the images by estimating a coaligned height map, which acts as a pixel-wise radial deformation field that orthorectifies each camera image to allow homographic registration. The height maps themselves are reparameterized as the output of an untrained encoder-decoder convolutional neural network (CNN) with the raw camera images as the input, which effectively removes many reconstruction artifacts. Our method also jointly estimates both the camera's dynamic 6D pose and its distortion using a nonparametric model, the latter of which is especially important in mesoscopic applications when using cameras not designed for imaging at short working distances, such as smartphone cameras. We also propose strategies for reducing computation time and memory, applicable to other multi-frame registration problems. Finally, we demonstrate our method using sequences of multi-megapixel images captured by an unstabilized smartphone on a variety of samples (e.g., painting brushstrokes, circuit board, seeds).
In this paper, we present a novel two-pass approach to unify streaming and non-streaming end-to-end (E2E) speech recognition in a single model. Our model adopts the hybrid CTC/attention architecture, in which the conformer layers in the encoder are modified. We propose a dynamic chunk-based attention strategy to allow arbitrary right context length. At inference time, the CTC decoder generates n-best hypotheses in a streaming way. The inference latency could be easily controlled by only changing the chunk size. The CTC hypotheses are then rescored by the attention decoder to get the final result. This efficient rescoring process causes very little sentence-level latency. Our experiments on the open 170-hour AISHELL-1 dataset show that, the proposed method can unify the streaming and non-streaming model simply and efficiently. On the AISHELL-1 test set, our unified model achieves 5.60% relative character error rate (CER) reduction in non-streaming ASR compared to a standard non-streaming transformer. The same model achieves 5.42% CER with 640ms latency in a streaming ASR system.
Encoder-decoder has been widely used in neural machine translation (NMT). A few methods have been proposed to improve it with multiple passes of decoding. However, their full potential is limited by a lack of appropriate termination policies. To address this issue, we present a novel architecture, Rewriter-Evaluator. It consists of a rewriter and an evaluator. Translating a source sentence involves multiple passes. At every pass, the rewriter produces a new translation to improve the past translation and the evaluator estimates the translation quality to decide whether to terminate the rewriting process. We also propose prioritized gradient descent (PGD) that facilitates training the rewriter and the evaluator jointly. Though incurring multiple passes of decoding, Rewriter-Evaluator with the proposed PGD method can be trained with a similar time to that of training encoder-decoder models. We apply the proposed architecture to improve the general NMT models (e.g., Transformer). We conduct extensive experiments on two translation tasks, Chinese-English and English-German, and show that the proposed architecture notably improves the performances of NMT models and significantly outperforms previous baselines.
Autoencoders are a class of artificial neural networks which have gained a lot of attention in the recent past. Using the encoder block of an autoencoder the input image can be compressed into a meaningful representation. Then a decoder is employed to reconstruct the compressed representation back to a version which looks like the input image. It has plenty of applications in the field of data compression and denoising. Another version of Autoencoders (AE) exist, called Variational AE (VAE) which acts as a generative model like GAN. Recently, an optimizer was introduced which is known as lookahead optimizer which significantly enhances the performances of Adam as well as SGD. In this paper, we implement Convolutional Autoencoders (CAE) and Convolutional Variational Autoencoders (CVAE) with lookahead optimizer (with Adam) and compare them with the Adam (only) optimizer counterparts. For this purpose, we have used a movie dataset comprising of natural images for the former case and CIFAR100 for the latter case. We show that lookahead optimizer (with Adam) improves the performance of CAEs for reconstruction of natural images.
Image captioning is a challenging computer vision task, which aims to generate a natural language description of an image. Most recent researches follow the encoder-decoder framework which depends heavily on the previous generated words for the current prediction. Such methods can not effectively take advantage of the future predicted information to learn complete semantics. In this paper, we propose Context-Aware Auxiliary Guidance (CAAG) mechanism that can guide the captioning model to perceive global contexts. Upon the captioning model, CAAG performs semantic attention that selectively concentrates on useful information of the global predictions to reproduce the current generation. To validate the adaptability of the method, we apply CAAG to three popular captioners and our proposal achieves competitive performance on the challenging Microsoft COCO image captioning benchmark, e.g. 132.2 CIDEr-D score on Karpathy split and 130.7 CIDEr-D (c40) score on official online evaluation server.
Robust semantic scene segmentation for automotive applications is a challenging problem in two key aspects: (1) labelling every individual scene pixel and (2) performing this task under unstable weather and illumination changes (e.g., foggy weather), which results in poor outdoor scene visibility. Such visibility limitations lead to non-optimal performance of generalised deep convolutional neural network-based semantic scene segmentation. In this paper, we propose an efficient end-to-end automotive semantic scene understanding approach that is robust to foggy weather conditions. As an end-to-end pipeline, our proposed approach provides: (1) the transformation of imagery from foggy to clear weather conditions using a domain transfer approach (correcting for poor visibility) and (2) semantically segmenting the scene using a competitive encoder-decoder architecture with low computational complexity (enabling real-time performance). Our approach incorporates RGB colour, depth and luminance images via distinct encoders with dense connectivity and features fusion to effectively exploit information from different inputs, which contributes to an optimal feature representation within the overall model. Using this architectural formulation with dense skip connections, our model achieves comparable performance to contemporary approaches at a fraction of the overall model complexity.
Texts appearing in daily scenes that can be recognized by OCR (Optical Character Recognition) tools contain significant information, such as street name, product brand and prices. Two tasks -- text-based visual question answering and text-based image captioning, with a text extension from existing vision-language applications, are catching on rapidly. To address these problems, many sophisticated multi-modality encoding frameworks (such as heterogeneous graph structure) are being used. In this paper, we argue that a simple attention mechanism can do the same or even better job without any bells and whistles. Under this mechanism, we simply split OCR token features into separate visual- and linguistic-attention branches, and send them to a popular Transformer decoder to generate answers or captions. Surprisingly, we find this simple baseline model is rather strong -- it consistently outperforms state-of-the-art (SOTA) models on two popular benchmarks, TextVQA and all three tasks of ST-VQA, although these SOTA models use far more complex encoding mechanisms. Transferring it to text-based image captioning, we also surpass the TextCaps Challenge 2020 winner. We wish this work to set the new baseline for this two OCR text related applications and to inspire new thinking of multi-modality encoder design. Code is available at https://github.com/ZephyrZhuQi/ssbaseline
This paper explores semi-supervised anomaly detection, a more practical setting for anomaly detection where a small additional set of labeled samples are provided. We propose a new KL-divergence based objective function for semi-supervised anomaly detection, and show that two factors: the mutual information between the data and latent representations, and the entropy of latent representations, constitute an integral objective function for anomaly detection. To resolve the contradiction in simultaneously optimizing the two factors, we propose a novel encoder-decoder-encoder structure, with the first encoder focusing on optimizing the mutual information and the second encoder focusing on optimizing the entropy. The two encoders are enforced to share similar encoding with a consistent constraint on their latent representations. Extensive experiments have revealed that the proposed method significantly outperforms several state-of-the-arts on multiple benchmark datasets, including medical diagnosis and several classic anomaly detection benchmarks.
As moving objects always draw more attention of human eyes, the temporal motive information is always exploited complementarily with spatial information to detect salient objects in videos. Although efficient tools such as optical flow have been proposed to extract temporal motive information, it often encounters difficulties when used for saliency detection due to the movement of camera or the partial movement of salient objects. In this paper, we investigate the complimentary roles of spatial and temporal information and propose a novel dynamic spatiotemporal network (DS-Net) for more effective fusion of spatiotemporal information. We construct a symmetric two-bypass network to explicitly extract spatial and temporal features. A dynamic weight generator (DWG) is designed to automatically learn the reliability of corresponding saliency branch. And a top-down cross attentive aggregation (CAA) procedure is designed so as to facilitate dynamic complementary aggregation of spatiotemporal features. Finally, the features are modified by spatial attention with the guidance of coarse saliency map and then go through decoder part for final saliency map. Experimental results on five benchmarks VOS, DAVIS, FBMS, SegTrack-v2, and ViSal demonstrate that the proposed method achieves superior performance than state-of-the-art algorithms. The source code is available at https://github.com/TJUMMG/DS-Net.
The successful emotional conversation system depends on sufficient perception and appropriate expression of emotions. In a real-life conversation, humans firstly instinctively perceive emotions from multi-source information, including the emotion flow hidden in dialogue history, facial expressions, audio, and personalities of speakers. Then, they convey suitable emotions according to their personalities, but these multiple types of information are insufficiently exploited in emotional conversation fields. To address this issue, in this paper, we propose a heterogeneous graph-based model for emotional conversation generation. Firstly, we design a Heterogeneous Graph-Based Encoder to represent the conversation content (i.e., the dialogue history, its emotion flow, facial expressions, audio, and speakers' personalities) with a heterogeneous graph neural network, and then predict suitable emotions for feedback. Secondly, we employ an Emotion-Personality-Aware Decoder to generate a response relevant to the conversation context as well as with appropriate emotions, through taking the encoded graph representations, the predicted emotions by the encoder and the personality of the current speaker as inputs. Experiments on both automatic and human evaluation show that our method can effectively perceive emotions from multi-source knowledge and generate a satisfactory response. Furthermore, based on the up-to-date text generator BART, our model still can achieve consistent improvement, which significantly outperforms some existing state-of-the-art models.
The ability to flexibly bind features into coherent wholes from different perspectives is a hallmark of cognition and intelligence. Importantly, the binding problem is not only relevant for vision but also for general intelligence, sensorimotor integration, event processing, and language. Various artificial neural network models have tackled this problem with dynamic neural fields and related approaches. Here we focus on a generative encoder-decoder architecture that adapts its perspective and binds features by means of retrospective inference. We first train a model to learn sufficiently accurate generative models of dynamic biological motion or other harmonic motion patterns, such as a pendulum. We then scramble the input to a certain extent, possibly vary the perspective onto it, and propagate the prediction error back onto a binding matrix, that is, hidden neural states that determine feature binding. Moreover, we propagate the error further back onto perspective taking neurons, which rotate and translate the input features onto a known frame of reference. Evaluations show that the resulting gradient-based inference process solves the perspective taking and binding problem for known biological motion patterns, essentially yielding a Gestalt perception mechanism. In addition, redundant feature properties and population encodings are shown to be highly useful. While we evaluate the algorithm on biological motion patterns, the principled approach should be applicable to binding and Gestalt perception problems in other domains.
We examine the role of character patterns in three tasks: morphological analysis, lemmatization and copy. We use a modified version of the standard sequence-to-sequence model, where the encoder is a pattern matching network. Each pattern scores all possible N character long subwords (substrings) on the source side, and the highest scoring subword's score is used to initialize the decoder as well as the input to the attention mechanism. This method allows learning which subwords of the input are important for generating the output. By training the models on the same source but different target, we can compare what subwords are important for different tasks and how they relate to each other. We define a similarity metric, a generalized form of the Jaccard similarity, and assign a similarity score to each pair of the three tasks that work on the same source but may differ in target. We examine how these three tasks are related to each other in 12 languages. Our code is publicly available.
Recently end-to-end scene text spotting has become a popular research topic due to its advantages of global optimization and high maintainability in real applications. Most methods attempt to develop various region of interest (RoI) operations to concatenate the detection part and the sequence recognition part into a two-stage text spotting framework. However, in such framework, the recognition part is highly sensitive to the detected results (e.g.), the compactness of text contours). To address this problem, in this paper, we propose a novel Mask AttentioN Guided One-stage text spotting framework named MANGO, in which character sequences can be directly recognized without RoI operation. Concretely, a position-aware mask attention module is developed to generate attention weights on each text instance and its characters. It allows different text instances in an image to be allocated on different feature map channels which are further grouped as a batch of instance features. Finally, a lightweight sequence decoder is applied to generate the character sequences. It is worth noting that MANGO inherently adapts to arbitrary-shaped text spotting and can be trained end-to-end with only coarse position information (e.g.), rectangular bounding box) and text annotations. Experimental results show that the proposed method achieves competitive and even new state-of-the-art performance on both regular and irregular text spotting benchmarks, i.e., ICDAR 2013, ICDAR 2015, Total-Text, and SCUT-CTW1500.
Automatic text summarization extracts important information from texts and presents the information in the form of a summary. Abstractive summarization approaches progressed significantly by switching to deep neural networks, but results are not yet satisfactory, especially for languages where large training sets do not exist. In several natural language processing tasks, a cross-lingual model transfer is successfully applied in less-resource languages. For summarization, the cross-lingual model transfer was not attempted due to a non-reusable decoder side of neural models that cannot correct target language generation. In our work, we use a pre-trained English summarization model based on deep neural networks and sequence-to-sequence architecture to summarize Slovene news articles. We address the problem of inadequate decoder by using an additional language model for the evaluation of the generated text in target language. We test several cross-lingual summarization models with different amounts of target data for fine-tuning. We assess the models with automatic evaluation measures and conduct a small-scale human evaluation. Automatic evaluation shows that the summaries of our best cross-lingual model are useful and of quality similar to the model trained only in the target language. Human evaluation shows that our best model generates summaries with high accuracy and acceptable readability. However, similar to other abstractive models, our models are not perfect and may occasionally produce misleading or absurd content.
Multi-step prediction is considered of major significance for time series analysis in many real life problems. Existing methods mainly focus on one-step-ahead forecasting, since multiple step forecasting generally fails due to accumulation of prediction errors. This paper presents a novel approach for predicting plant growth in agriculture, focusing on prediction of plant Stem Diameter Variations (SDV). The proposed approach consists of three main steps. At first, wavelet decomposition is applied to the original data, as to facilitate model fitting and reduce noise in them. Then an encoder-decoder framework is developed using Long Short Term Memory (LSTM) and used for appropriate feature extraction from the data. Finally, a recurrent neural network including LSTM and an attention mechanism is proposed for modelling long-term dependencies in the time series data. Experimental results are presented which illustrate the good performance of the proposed approach and that it significantly outperforms the existing models, in terms of error criteria such as RMSE, MAE and MAPE.
Adversarial robustness of deep neural networks is an extensively studied problem in the literature and various methods have been proposed to defend against adversarial images. However, only a handful of defense methods have been developed for defending against attacked videos. In this paper, we propose a novel Over-and-Under complete restoration network for Defending against adversarial videos (OUDefend). Most restoration networks adopt an encoder-decoder architecture that first shrinks spatial dimension then expands it back. This approach learns undercomplete representations, which have large receptive fields to collect global information but overlooks local details. On the other hand, overcomplete representations have opposite properties. Hence, OUDefend is designed to balance local and global features by learning those two representations. We attach OUDefend to target video recognition models as a feature restoration block and train the entire network end-to-end. Experimental results show that the defenses focusing on images may be ineffective to videos, while OUDefend enhances robustness against different types of adversarial videos, ranging from additive attacks, multiplicative attacks to physically realizable attacks. Code: https://github.com/shaoyuanlo/OUDefend
Over the last few years, the performance of inpainting to fill missing regions has shown significant improvements by using deep neural networks. Most of inpainting work create a visually plausible structure and texture, however, due to them often generating a blurry result, final outcomes appear unrealistic and make feel heterogeneity. In order to solve this problem, the existing methods have used a patch based solution with deep neural network, however, these methods also cannot transfer the texture properly. Motivated by these observation, we propose a patch based method. Texture Transform Attention network(TTA-Net) that better produces the missing region inpainting with fine details. The task is a single refinement network and takes the form of U-Net architecture that transfers fine texture features of encoder to coarse semantic features of decoder through skip-connection. Texture Transform Attention is used to create a new reassembled texture map using fine textures and coarse semantics that can efficiently transfer texture information as a result. To stabilize training process, we use a VGG feature layer of ground truth and patch discriminator. We evaluate our model end-to-end with the publicly available datasets CelebA-HQ and Places2 and demonstrate that images of higher quality can be obtained to the existing state-of-the-art approaches.
Dimensionality reduction methods for count data are critical to a wide range of applications in medical informatics and other fields where model interpretability is paramount. For such data, hierarchical Poisson matrix factorization (HPF) and other sparse probabilistic non-negative matrix factorization (NMF) methods are considered to be interpretable generative models. They consist of sparse transformations for decoding their learned representations into predictions. However, sparsity in representation decoding does not necessarily imply sparsity in the encoding of representations from the original data features. HPF is often incorrectly interpreted in the literature as if it possesses encoder sparsity. The distinction between decoder sparsity and encoder sparsity is subtle but important. Due to the lack of encoder sparsity, HPF does not possess the column-clustering property of classical NMF -- the factor loading matrix does not sufficiently define how each factor is formed from the original features. We address this deficiency by self-consistently enforcing encoder sparsity, using a generalized additive model (GAM), thereby allowing one to relate each representation coordinate to a subset of the original data features. In doing so, the method also gains the ability to perform feature selection. We demonstrate our method on simulated data and give an example of how encoder sparsity is of practical use in a concrete application of representing inpatient comorbidities in Medicare patients.
Many speech synthesis datasets, especially those derived from audiobooks, naturally comprise sequences of utterances. Nevertheless, such data are commonly treated as individual, unordered utterances both when training a model and at inference time. This discards important prosodic phenomena above the utterance level. In this paper, we leverage the sequential nature of the data using an acoustic context encoder that produces an embedding of the previous utterance audio. This is input to the decoder in a Tacotron 2 model. The embedding is also used for a secondary task, providing additional supervision. We compare two secondary tasks: predicting the ordering of utterance pairs, and predicting the embedding of the current utterance audio. Results show that the relation between consecutive utterances is informative: our proposed model significantly improves naturalness over a Tacotron 2 baseline.