Neural dialogue generation models trained with the one-hot target distribution suffer from the over-confidence issue, which leads to poor generation diversity as widely reported in the literature. Although existing approaches such as label smoothing can alleviate this issue, they fail to adapt to diverse dialog contexts. In this paper, we propose an Adaptive Label Smoothing (AdaLabel) approach that can adaptively estimate a target label distribution at each time step for different contexts. The maximum probability in the predicted distribution is used to modify the soft target distribution produced by a novel light-weight bi-directional decoder module. The resulting target distribution is aware of both previous and future contexts and is adjusted to avoid over-training the dialogue model. Our model can be trained in an end-to-end manner. Extensive experiments on two benchmark datasets show that our approach outperforms various competitive baselines in producing diverse responses.
Automatically generating medical reports for retinal images is one of the promising ways to help ophthalmologists reduce their workload and improve work efficiency. In this work, we propose a new context-driven encoding network to automatically generate medical reports for retinal images. The proposed model is mainly composed of a multi-modal input encoder and a fused-feature decoder. Our experimental results show that our proposed method is capable of effectively leveraging the interactive information between the input image and context, i.e., keywords in our case. The proposed method creates more accurate and meaningful reports for retinal images than baseline models and achieves state-of-the-art performance. This performance is shown in several commonly used metrics for the medical report generation task: BLEU-avg (+16%), CIDEr (+10.2%), and ROUGE (+8.6%).
Perception, in theoretical neuroscience, has been modeled as the encoding of external stimuli into internal signals, which are then decoded. The Bayesian mean is an important decoder, as it is optimal for purposes of both estimation and discrimination. We present widely-applicable approximations to the bias and to the variance of the Bayesian mean, obtained under the minimal and biologically-relevant assumption that the encoding results from a series of independent, though not necessarily identically-distributed, signals. Simulations substantiate the accuracy of our approximations in the small-noise regime. The bias of the Bayesian mean comprises two components: one driven by the prior, and one driven by the precision of the encoding. If the encoding is 'efficient', the two components have opposite effects; their relative strengths are determined by the objective that the encoding optimizes. The experimental literature on perception reports both 'Bayesian' biases directed towards prior expectations, and opposite, 'anti-Bayesian' biases. We show that different tasks are indeed predicted to yield such contradictory biases, under a consistently-optimal encoding-decoding model. Moreover, we recover Wei and Stocker's "law of human perception", a relation between the bias of the Bayesian mean and the derivative of its variance, and show how the coefficient of proportionality in this law depends on the task at hand. Our results provide a parsimonious theory of optimal perception under constraints, in which encoding and decoding are adapted both to the prior and to the task faced by the observer.
Parallel cross-lingual summarization data is scarce, requiring models to better use the limited available cross-lingual resources. Existing methods to do so often adopt sequence-to-sequence networks with multi-task frameworks. Such approaches apply multiple decoders, each of which is utilized for a specific task. However, these independent decoders share no parameters, hence fail to capture the relationships between the discrete phrases of summaries in different languages, breaking the connections in order to transfer the knowledge of the high-resource languages to low-resource languages. To bridge these connections, we propose a novel Multi-Task framework for Cross-Lingual Abstractive Summarization (MCLAS) in a low-resource setting. Employing one unified decoder to generate the sequential concatenation of monolingual and cross-lingual summaries, MCLAS makes the monolingual summarization task a prerequisite of the cross-lingual summarization (CLS) task. In this way, the shared decoder learns interactions involving alignments and summary patterns across languages, which encourages attaining knowledge transfer. Experiments on two CLS datasets demonstrate that our model significantly outperforms three baseline models in both low-resource and full-dataset scenarios. Moreover, in-depth analysis on the generated summaries and attention heads verifies that interactions are learned well using MCLAS, which benefits the CLS task under limited parallel resources.
Image captioning generates text that describes scenes from input images. It has been developed for high quality images taken in clear weather. However, in bad weather conditions, such as heavy rain, snow, and dense fog, the poor visibility owing to rain streaks, rain accumulation, and snowflakes causes a serious degradation of image quality. This hinders the extraction of useful visual features and results in deteriorated image captioning performance. To address practical issues, this study introduces a new encoder for captioning heavy rain images. The central idea is to transform output features extracted from heavy rain input images into semantic visual features associated with words and sentence context. To achieve this, a target encoder is initially trained in an encoder-decoder framework to associate visual features with semantic words. Subsequently, the objects in a heavy rain image are rendered visible by using an initial reconstruction subnetwork (IRS) based on a heavy rain model. The IRS is then combined with another semantic visual feature matching subnetwork (SVFMS) to match the output features of the IRS with the semantic visual features of the pretrained target encoder. The proposed encoder is based on the joint learning of the IRS and SVFMS. It is is trained in an end-to-end manner, and then connected to the pretrained decoder for image captioning. It is experimentally demonstrated that the proposed encoder can generate semantic visual features associated with words even from heavy rain images, thereby increasing the accuracy of the generated captions.
Many existing conversation models that are based on the encoder-decoder framework have focused on ways to make the encoder more complicated to enrich the context vectors so as to increase the diversity and informativeness of generated responses. However, these approaches face two problems. First, the decoder is too simple to effectively utilize the previously generated information and tends to generate duplicated and self-contradicting responses. Second, the complex encoder tends to generate diverse but incoherent responses because the complex context vectors may deviate from the original semantics of context. In this work, we proposed a conversation model named "THINK" (Teamwork generation Hover around Impressive Noticeable Keywords) to make the decoder more complicated and avoid generating duplicated and self-contradicting responses. The model simplifies the context vectors and increases the coherence of generated responses in a reasonable way. For this model, we propose Teamwork generation framework and Semantics Extractor. Compared with other baselines, both automatic and human evaluation showed the advantages of our model.
In this paper, we present our solution for the {\it IJCAI--PRICAI--20 3D AI Challenge: 3D Object Reconstruction from A Single Image}. We develop a variant of AtlasNet that consumes single 2D images and generates 3D point clouds through 2D to 3D mapping. To push the performance to the limit and present guidance on crucial implementation choices, we conduct extensive experiments to analyze the influence of decoder design and different settings on the normalization, projection, and sampling methods. Our method achieves 2nd place in the final track with a score of $70.88$, a chamfer distance of $36.87$, and a mean f-score of $59.18$. The source code of our method will be available at https://github.com/em-data/Enhanced_AtlasNet_3DReconstruction.
In recent years, significant progress has been made in the research of facial landmark detection. However, few prior works have thoroughly discussed about models for practical applications. Instead, they often focus on improving a couple of issues at a time while ignoring the others. To bridge this gap, we aim to explore a practical model that is accurate, robust, efficient, generalizable, and end-to-end trainable at the same time. To this end, we first propose a baseline model equipped with one transformer decoder as detection head. In order to achieve a better accuracy, we further propose two lightweight modules, namely dynamic query initialization (DQInit) and query-aware memory (QAMem). Specifically, DQInit dynamically initializes the queries of decoder from the inputs, enabling the model to achieve as good accuracy as the ones with multiple decoder layers. QAMem is designed to enhance the discriminative ability of queries on low-resolution feature maps by assigning separate memory values to each query rather than a shared one. With the help of QAMem, our model removes the dependence on high-resolution feature maps and is still able to obtain superior accuracy. Extensive experiments and analysis on three popular benchmarks show the effectiveness and practical advantages of the proposed model. Notably, our model achieves new state of the art on WFLW as well as competitive results on 300W and COFW, while still running at 50+ FPS.
Recently, opinion summarization, which is the generation of a summary from multiple reviews, has been conducted in a self-supervised manner by considering a sampled review as a pseudo summary. However, non-text data such as image and metadata related to reviews have been considered less often. To use the abundant information contained in non-text data, we propose a self-supervised multimodal opinion summarization framework called MultimodalSum. Our framework obtains a representation of each modality using a separate encoder for each modality, and the text decoder generates a summary. To resolve the inherent heterogeneity of multimodal data, we propose a multimodal training pipeline. We first pretrain the text encoder--decoder based solely on text modality data. Subsequently, we pretrain the non-text modality encoders by considering the pretrained text decoder as a pivot for the homogeneous representation of multimodal data. Finally, to fuse multimodal representations, we train the entire framework in an end-to-end manner. We demonstrate the superiority of MultimodalSum by conducting experiments on Yelp and Amazon datasets.
Most consumer-grade digital cameras can only capture a limited range of luminance in real-world scenes due to sensor constraints. Besides, noise and quantization errors are often introduced in the imaging process. In order to obtain high dynamic range (HDR) images with excellent visual quality, the most common solution is to combine multiple images with different exposures. However, it is not always feasible to obtain multiple images of the same scene and most HDR reconstruction methods ignore the noise and quantization loss. In this work, we propose a novel learning-based approach using a spatially dynamic encoder-decoder network, HDRUNet, to learn an end-to-end mapping for single image HDR reconstruction with denoising and dequantization. The network consists of a UNet-style base network to make full use of the hierarchical multi-scale information, a condition network to perform pattern-specific modulation and a weighting network for selectively retaining information. Moreover, we propose a Tanh_L1 loss function to balance the impact of over-exposed values and well-exposed values on the network learning. Our method achieves the state-of-the-art performance in quantitative comparisons and visual quality. The proposed HDRUNet model won the second place in the single frame track of NITRE2021 High Dynamic Range Challenge.
For real-time semantic segmentation, how to increase the speed while maintaining high resolution is a problem that has been discussed and solved. Backbone design and fusion design have always been two essential parts of real-time semantic segmentation. We hope to design a light-weight network based on previous design experience and reach the level of state-of-the-art real-time semantic segmentation without any pre-training. To achieve this goal, a encoder-decoder architectures are proposed to solve this problem by applying a decoder network onto a backbone model designed for real-time segmentation tasks and designed three different ways to fuse semantics and detailed information in the aggregation phase. We have conducted extensive experiments on two semantic segmentation benchmarks. Experiments on the Cityscapes and CamVid datasets show that the proposed FRFNet strikes a balance between speed calculation and accuracy. It achieves 72% Mean Intersection over Union (mIoU%) on the Cityscapes test dataset with the speed of 144 on a single RTX 1080Ti card. The Code is available at https://github.com/favoMJ/FRFNet.
Deep learning has demonstrated significant improvements in medical image segmentation using a sufficiently large amount of training data with manual labels. Acquiring well-representative labels requires expert knowledge and exhaustive labors. In this paper, we aim to boost the performance of semi-supervised learning for medical image segmentation with limited labels using a self-ensembling contrastive learning technique. To this end, we propose to train an encoder-decoder network at image-level with small amounts of labeled images, and more importantly, we learn latent representations directly at feature-level by imposing contrastive loss on unlabeled images. This method strengthens intra-class compactness and inter-class separability, so as to get a better pixel classifier. Moreover, we devise a student encoder for online learning and an exponential moving average version of it, called teacher encoder, to improve the performance iteratively in a self-ensembling manner. To construct contrastive samples with unlabeled images, two sampling strategies that exploit structure similarity across medical images and utilize pseudo-labels for construction, termed region-aware and anatomical-aware contrastive sampling, are investigated. We conduct extensive experiments on an MRI and a CT segmentation dataset and demonstrate that in a limited label setting, the proposed method achieves state-of-the-art performance. Moreover, the anatomical-aware strategy that prepares contrastive samples on-the-fly using pseudo-labels realizes better contrastive regularization on feature representations.
Despite existing pioneering works on sign language translation (SLT), there is a non-trivial obstacle, i.e., the limited quantity of parallel sign-text data. To tackle this parallel data bottleneck, we propose a sign back-translation (SignBT) approach, which incorporates massive spoken language texts into SLT training. With a text-to-gloss translation model, we first back-translate the monolingual text to its gloss sequence. Then, the paired sign sequence is generated by splicing pieces from an estimated gloss-to-sign bank at the feature level. Finally, the synthetic parallel data serves as a strong supplement for the end-to-end training of the encoder-decoder SLT framework. To promote the SLT research, we further contribute CSL-Daily, a large-scale continuous SLT dataset. It provides both spoken language translations and gloss-level annotations. The topic revolves around people's daily lives (e.g., travel, shopping, medical care), the most likely SLT application scenario. Extensive experimental results and analysis of SLT methods are reported on CSL-Daily. With the proposed sign back-translation method, we obtain a substantial improvement over previous state-of-the-art SLT methods.
In this paper, we propose a new deep image compression framework called Complexity and Bitrate Adaptive Network (CBANet), which aims to learn one single network to support variable bitrate coding under different computational complexity constraints. In contrast to the existing state-of-the-art learning based image compression frameworks that only consider the rate-distortion trade-off without introducing any constraint related to the computational complexity, our CBANet considers the trade-off between the rate and distortion under dynamic computational complexity constraints. Specifically, to decode the images with one single decoder under various computational complexity constraints, we propose a new multi-branch complexity adaptive module, in which each branch only takes a small portion of the computational budget of the decoder. The reconstructed images with different visual qualities can be readily generated by using different numbers of branches. Furthermore, to achieve variable bitrate decoding with one single decoder, we propose a bitrate adaptive module to project the representation from a base bitrate to the expected representation at a target bitrate for transmission. Then it will project the transmitted representation at the target bitrate back to that at the base bitrate for the decoding process. The proposed bit adaptive module can significantly reduce the storage requirement for deployment platforms. As a result, our CBANet enables one single codec to support multiple bitrate decoding under various computational complexity constraints. Comprehensive experiments on two benchmark datasets demonstrate the effectiveness of our CBANet for deep image compression.
In recent years the interest in segmentation has been growing, being used in a wide range of applications such as fraud detection, anomaly detection in public health and intrusion detection. We present an ablation study of FgSegNet_v2, analysing its three stages: (i) Encoder, (ii) Feature Pooling Module and (iii) Decoder. The result of this study is a proposal of a variation of the aforementioned method that surpasses state of the art results. Three datasets are used for testing: CDNet2014, SBI2015 and CityScapes. In CDNet2014 we got an overall improvement compared to the state of the art, mainly in the LowFrameRate subset. The presented approach is promising as it produces comparable results with the state of the art (SBI2015 and Cityscapes datasets) in very different conditions, such as different lighting conditions.
To achieve higher coding efficiency, Versatile Video Coding (VVC) includes several novel components, but at the expense of increasing decoder computational complexity. These technologies at a low bit rate often create contouring and ringing effects on the reconstructed frames and introduce various blocking artifacts at block boundaries. To suppress those visual artifacts, the VVC framework supports four post-processing filter operations. The interoperation of these filters introduces extra signaling bits and eventually becomes overhead at higher resolution video processing. In this paper, a novel deep learning-based model is proposed for sample adaptive offset (SAO) nonlinear filtering operation and substantiated the merits of intra-inter frame quality enhancement. We introduced a variable filter size multi-scale CNN (MSCNN) to improve the denoising operation and incorporated strided deconvolution for further computation improvement. We demonstrated that our deconvolution model can effectively be trained by leveraging the high-frequency edge features learned in a parallel fashion using feature fusion and residual learning. The simulation results demonstrate that the proposed method outperforms the baseline VVC method in BD-BR, BD-PSNR measurements and achieves an average of 3.762 % bit rate saving on the standard video test sequences.
Hierarchical structures are popular in recent vision transformers, however, they require sophisticated designs and massive datasets to work well. In this paper, we explore the idea of nesting basic local transformers on non-overlapping image blocks and aggregating them in a hierarchical way. We find that the block aggregation function plays a critical role in enabling cross-block non-local information communication. This observation leads us to design a simplified architecture that requires minor code changes upon the original vision transformer. The benefits of the proposed judiciously-selected design are threefold: (1) NesT converges faster and requires much less training data to achieve good generalization on both ImageNet and small datasets like CIFAR; (2) when extending our key ideas to image generation, NesT leads to a strong decoder that is 8$\times$ faster than previous transformer-based generators; and (3) we show that decoupling the feature learning and abstraction processes via this nested hierarchy in our design enables constructing a novel method (named GradCAT) for visually interpreting the learned model. Source code is available https://github.com/google-research/nested-transformer.
Abstractive summarization for long-document or multi-document remains challenging for the Seq2Seq architecture, as Seq2Seq is not good at analyzing long-distance relations in text. In this paper, we present BASS, a novel framework for Boosting Abstractive Summarization based on a unified Semantic graph, which aggregates co-referent phrases distributing across a long range of context and conveys rich relations between phrases. Further, a graph-based encoder-decoder model is proposed to improve both the document representation and summary generation process by leveraging the graph structure. Specifically, several graph augmentation methods are designed to encode both the explicit and implicit relations in the text while the graph-propagation attention mechanism is developed in the decoder to select salient content into the summary. Empirical results show that the proposed architecture brings substantial improvements for both long-document and multi-document summarization tasks.
For a computer to naturally interact with a human, it needs to be human-like. In this paper, we propose a neural response generation model with multi-task learning of generation and classification, focusing on emotion. Our model based on BART (Lewis et al., 2020), a pre-trained transformer encoder-decoder model, is trained to generate responses and recognize emotions simultaneously. Furthermore, we weight the losses for the tasks to control the update of parameters. Automatic evaluations and crowdsourced manual evaluations show that the proposed model makes generated responses more emotionally aware.
Recent achievements in end-to-end deep learning have encouraged the exploration of tasks dealing with highly structured data with unified deep network models. Having such models for compressing audio signals has been challenging since it requires discrete representations that are not easy to train with end-to-end backpropagation. In this paper, we present an end-to-end deep learning approach that combines recurrent neural networks (RNNs) within the training strategy of variational autoencoders (VAEs) with a binary representation of the latent space. We apply a reparametrization trick for the Bernoulli distribution for the discrete representations, which allows smooth backpropagation. In addition, our approach allows the separation of the encoder and decoder, which is necessary for compression tasks. To our best knowledge, this is the first end-to-end learning for a single audio compression model with RNNs, and our model achieves a Signal to Distortion Ratio (SDR) of 20.54.