Models, code, and papers for "Zhengyang Wang":
Dilated convolutions, also known as atrous convolutions, have been widely explored in deep convolutional neural networks (DCNNs) for various tasks like semantic image segmentation, object detection, audio generation, video modeling, and machine translation. However, dilated convolutions suffer from the gridding artifacts, which hampers the performance of DCNNs with dilated convolutions. In this work, we propose two simple yet effective degridding methods by studying a decomposition of dilated convolutions. Unlike existing models, which explore solutions by focusing on a block of cascaded dilated convolutional layers, our methods address the gridding artifacts by smoothing the dilated convolution itself. By analyzing them in both the original operation and the decomposition views, we further point out that the two degridding approaches are intrinsically related and define separable and shared (SS) operations, which generalize the proposed methods. We evaluate our methods thoroughly on two datasets and visualize the smoothing effect through effective receptive field analysis. Experimental results show that our methods yield significant and consistent improvements on the performance of DCNNs with dilated convolutions, while adding negligible amounts of extra training parameters.
Visual question answering is a recently proposed artificial intelligence task that requires a deep understanding of both images and texts. In deep learning, images are typically modeled through convolutional neural networks, and texts are typically modeled through recurrent neural networks. While the requirement for modeling images is similar to traditional computer vision tasks, such as object recognition and image classification, visual question answering raises a different need for textual representation as compared to other natural language processing tasks. In this work, we perform a detailed analysis on natural language questions in visual question answering. Based on the analysis, we propose to rely on convolutional neural networks for learning textual representations. By exploring the various properties of convolutional neural networks specialized for text data, such as width and depth, we present our "CNN Inception + Gate" model. We show that our model improves question representations and thus the overall accuracy of visual question answering models. We also show that the text representation requirement in visual question answering is more complicated and comprehensive than that in conventional natural language processing tasks, making it a better task to evaluate textual representation methods. Shallow models like fastText, which can obtain comparable results with deep learning models in tasks like text classification, are not suitable in visual question answering.
Convolutional neural networks (CNNs) have shown great capability of solving various artificial intelligence tasks. However, the increasing model size has raised challenges in employing them in resource-limited applications. In this work, we propose to compress deep models by using channel-wise convolutions, which re- place dense connections among feature maps with sparse ones in CNNs. Based on this novel operation, we build light-weight CNNs known as ChannelNets. Channel- Nets use three instances of channel-wise convolutions; namely group channel-wise convolutions, depth-wise separable channel-wise convolutions, and the convolu- tional classification layer. Compared to prior CNNs designed for mobile devices, ChannelNets achieve a significant reduction in terms of the number of parameters and computational cost without loss in accuracy. Notably, our work represents the first attempt to compress the fully-connected classification layer, which usually accounts for about 25% of total parameters in compact CNNs. Experimental results on the ImageNet dataset demonstrate that ChannelNets achieve consistently better performance compared to prior methods.
Convolutional neural networks (CNNs) have achieved great success on grid-like data such as images, but face tremendous challenges in learning from more generic data such as graphs. In CNNs, the trainable local filters enable the automatic extraction of high-level features. The computation with filters requires a fixed number of ordered units in the receptive fields. However, the number of neighboring units is neither fixed nor are they ordered in generic graphs, thereby hindering the applications of convolutional operations. Here, we address these challenges by proposing the learnable graph convolutional layer (LGCL). LGCL automatically selects a fixed number of neighboring nodes for each feature based on value ranking in order to transform graph data into grid-like structures in 1-D format, thereby enabling the use of regular convolutional operations on generic graphs. To enable model training on large-scale graphs, we propose a sub-graph training method to reduce the excessive memory and computational resource requirements suffered by prior methods on graph convolutions. Our experimental results on node classification tasks in both transductive and inductive learning settings demonstrate that our methods can achieve consistently better performance on the Cora, Citeseer, Pubmed citation network, and protein-protein interaction network datasets. Our results also indicate that the proposed methods using sub-graph training strategy are more efficient as compared to prior approaches.
The key idea of variational auto-encoders (VAEs) resembles that of traditional auto-encoder models in which spatial information is supposed to be explicitly encoded in the latent space. However, the latent variables in VAEs are vectors, which are commonly interpreted as multiple feature maps of size 1x1. Such representations can only convey spatial information implicitly when coupled with powerful decoders. In this work, we propose spatial VAEs that use latent variables as feature maps of larger size to explicitly capture spatial information. This is achieved by allowing the latent variables to be sampled from matrix-variate normal (MVN) distributions whose parameters are computed from the encoder network. To increase dependencies among locations on latent feature maps and reduce the number of parameters, we further propose spatial VAEs via low-rank MVN distributions. Experimental results show that the proposed spatial VAEs outperform original VAEs in capturing rich structural and spatial information.
Visualizing the details of different cellular structures is of great importance to elucidate cellular functions. However, it is challenging to obtain high quality images of different structures directly due to complex cellular environments. Fluorescence microscopy is a popular technique to label different structures but has several drawbacks. In particular, labeling is time consuming and may affect cell morphology, and simultaneous labels are inherently limited. This raises the need of building computational models to learn relationships between unlabeled and labeled fluorescence images, and to infer fluorescent labels of other unlabeled fluorescence images. We propose to develop a novel deep model for fluorescence image prediction. We first propose a novel network layer, known as the global transformer layer, that fuses global information from inputs effectively. The proposed global transformer layer can generate outputs with arbitrary dimensions, and can be employed for all the regular, down-sampling, and up-sampling operators. We then incorporate our proposed global transformer layers and dense blocks to build an U-Net like network. We believe such a design can promote feature reusing between layers. In addition, we propose a multi-scale input strategy to encourage networks to capture features at different scales. We conduct evaluations across various label-free prediction tasks to demonstrate the effectiveness of our approach. Both quantitative and qualitative results show that our method outperforms the state-of-the-art approach significantly. It is also shown that our proposed global transformer layer is useful to improve the fluorescence image prediction results.
An important step in early brain development study is to perform automatic segmentation of infant brain magnetic resonance (MR) images into cerebrospinal fluid (CSF), gray matter (GM) and white matter (WM) regions. This task is especially challenging in the isointense stage (approximately 6-8 months of age) when GM and WM exhibit similar levels of intensities in MR images. Deep learning has shown its great promise in various image segmentation tasks. However, existing models do not have an efficient and effective way to aggregate global information. They also suffer from information loss during up-sampling operations. In this work, we address these problems by proposing a global aggregation block, which can be flexibly used for global information fusion. We build a novel model based on 3D U-Net to make fast and accurate voxel-wise dense prediction. We perform thorough experiments, and results indicate that our model outperforms previous best models significantly on 3D multimodality isointense infant brain MR image segmentation.
Deconvolutional layers have been widely used in a variety of deep models for up-sampling, including encoder-decoder networks for semantic segmentation and deep generative models for unsupervised learning. One of the key limitations of deconvolutional operations is that they result in the so-called checkerboard problem. This is caused by the fact that no direct relationship exists among adjacent pixels on the output feature map. To address this problem, we propose the pixel deconvolutional layer (PixelDCL) to establish direct relationships among adjacent pixels on the up-sampled feature map. Our method is based on a fresh interpretation of the regular deconvolution operation. The resulting PixelDCL can be used to replace any deconvolutional layer in a plug-and-play manner without compromising the fully trainable capabilities of original models. The proposed PixelDCL may result in slight decrease in efficiency, but this can be overcome by an implementation trick. Experimental results on semantic segmentation demonstrate that PixelDCL can consider spatial features such as edges and shapes and yields more accurate segmentation outputs than deconvolutional layers. When used in image generation tasks, our PixelDCL can largely overcome the checkerboard problem suffered by regular deconvolution operations.
This paper proposes a multi-level feature learning framework for human action recognition using a single body-worn inertial sensor. The framework consists of three phases, respectively designed to analyze signal-based (low-level), components (mid-level) and semantic (high-level) information. Low-level features capture the time and frequency domain property while mid-level representations learn the composition of the action. The Max-margin Latent Pattern Learning (MLPL) method is proposed to learn high-level semantic descriptions of latent action patterns as the output of our framework. The proposed method achieves the state-of-the-art performances, 88.7%, 98.8% and 72.6% (weighted F1 score) respectively, on Skoda, WISDM and OPP datasets.
In this paper, we propose an innovative end-to-end subtitle detection and recognition system for videos in East Asian languages. Our end-to-end system consists of multiple stages. Subtitles are firstly detected by a novel image operator based on the sequence information of consecutive video frames. Then, an ensemble of Convolutional Neural Networks (CNNs) trained on synthetic data is adopted for detecting and recognizing East Asian characters. Finally, a dynamic programming approach leveraging language models is applied to constitute results of the entire body of text lines. The proposed system achieves average end-to-end accuracies of 98.2% and 98.3% on 40 videos in Simplified Chinese and 40 videos in Traditional Chinese respectively, which is a significant outperformance of other existing methods. The near-perfect accuracy of our system dramatically narrows the gap between human cognitive ability and state-of-the-art algorithms used for such a task.