Enhua Wu

Bio: Enhua Wu is an academic researcher from Chinese Academy of Sciences. The author has contributed to research in topics: Rendering (computer graphics) & Polygon mesh. The author has an hindex of 24, co-authored 266 publications receiving 10340 citations. Previous affiliations of Enhua Wu include University of Macau & Academia Sinica.

FreePipe: a programmable parallel rendering architecture for efficient multi-fragment effects

Abstract: In the past decade, modern GPUs have provided increasing programmability with vertex, geometry and fragment shaders. However, many classical problems have not been efficiently solved using the current graphics pipeline where some stages are still fixed functions on chip. In particular, multi-fragment effects, especially order-independent transparency, require programmability of the blending stage, that makes it difficult to be solved in a single geometry pass. In this paper we present FreePipe, a system for programmable parallel rendering that can run entirely on current graphics hardware and has performance comparable with the traditional graphics pipeline. Within this framework, two schemes for the efficient rendering of multi-fragment effects in a single geometry pass have been developed by exploiting CUDA atomic operations. Both schemes have achieved significant speedups compared to the state-of-the-art methods that are based on traditional graphics pipelines.

Robust dense reconstruction by range merging based on confidence estimation

Abstract: Although the stereo matching problem has been extensively studied during the past decades, automatically computing a dense 3D reconstruction from several multiple views is still a difficult task owing to the problems of textureless regions, outliers, detail loss, and various other factors. In this paper, these difficult problems are handled effectively by a robust model that outputs an accurate and dense reconstruction as the final result from an input of multiple images captured by a normal camera. First, the positions of the camera and sparse 3D points are estimated by a structure-from-motion algorithm and we compute the range map with a confidence estimation for each image in our approach. Then all the range maps are integrated into a fine point cloud data set. In the final step we use a Poisson reconstruction algorithm to finish the reconstruction. The major contributions of the work lie in the following points: effective range-computation and confidence-estimation methods are proposed to handle the problems of textureless regions, outliers and detail loss. Then, the range maps are merged into the point cloud data in terms of a confidence-estimation. Finally, Poisson reconstruction algorithm completes the dense mesh. In addition, texture mapping is also implemented as a post-processing work for obtaining good visual effects. Experimental results are presented to demonstrate the effectiveness of the proposed approach.

Multi-layer depth peeling via fragment sort

Abstract: We present an accelerated depth peeling algorithm for order-independent transparency rendering on graphics hardware. Unlike traditional depth peeling which only peels one layer of transparent pixels per rendering pass, our algorithm peels multiple layers simultaneously per rendering pass. Our acceleration is achieved via our fragment program which sorts and writes multiple fragment colors and depths via MRT. A notable feature of our algorithm is that it is robust against the unreliable parallel read-after-write behavior in current graphics hardware, guaranteeing correct transparency ordering. For ordinary scenes rendered under RGBA8 color precision, we achieve up to 8x speed-up over conventional depth peeling with current generation graphics hardware. Our algorithm is simple to implement on current GPU without any hardware modification. In addition, it does not require applications to perform any pre-sorting of transparent geometry.

FreePipe: a programmable parallel rendering architecture for efficient multi-fragment effects.

Abstract: In the past decade, modern GPUs have provided increasing programmability with vertex, geometry and fragment shaders. However, many classical problems have not been efficiently solved using the current graphics pipeline where some stages are still fixed functions on chip. In particular, multi-fragment effects, especially order-independent transparency, require programmability of the blending stage, that makes it difficult to be solved in a single geometry pass. In this paper we present FreePipe, a system for programmable parallel rendering that can run entirely on current graphics hardware and has performance comparable with the traditional graphics pipeline. Within this framework, two schemes for the efficient rendering of multi-fragment effects in a single geometry pass have been developed by exploiting CUDA atomic operations. Both schemes have achieved significant speedups compared to the state-of-the-art methods that are based on traditional graphics pipelines.

Multi-Level Discriminative Dictionary Learning With Application to Large Scale Image Classification

Abstract: The sparse coding technique has shown flexibility and capability in image representation and analysis. It is a powerful tool in many visual applications. Some recent work has shown that incorporating the properties of task (such as discrimination for classification task) into dictionary learning is effective for improving the accuracy. However, the traditional supervised dictionary learning methods suffer from high computation complexity when dealing with large number of categories, making them less satisfactory in large scale applications. In this paper, we propose a novel multi-level discriminative dictionary learning method and apply it to large scale image classification. Our method takes advantage of hierarchical category correlation to encode multi-level discriminative information. Each internal node of the category hierarchy is associated with a discriminative dictionary and a classification model. The dictionaries at different layers are learnt to capture the information of different scales. Moreover, each node at lower layers also inherits the dictionary of its parent, so that the categories at lower layers can be described with multi-scale information. The learning of dictionaries and associated classification models is jointly conducted by minimizing an overall tree loss. The experimental results on challenging data sets demonstrate that our approach achieves excellent accuracy and competitive computation cost compared with other sparse coding methods for large scale image classification.

Squeeze-and-Excitation Networks

Abstract: The central building block of convolutional neural networks (CNNs) is the convolution operator, which enables networks to construct informative features by fusing both spatial and channel-wise information within local receptive fields at each layer. A broad range of prior research has investigated the spatial component of this relationship, seeking to strengthen the representational power of a CNN by enhancing the quality of spatial encodings throughout its feature hierarchy. In this work, we focus instead on the channel relationship and propose a novel architectural unit, which we term the “Squeeze-and-Excitation” (SE) block, that adaptively recalibrates channel-wise feature responses by explicitly modelling interdependencies between channels. We show that these blocks can be stacked together to form SENet architectures that generalise extremely effectively across different datasets. We further demonstrate that SE blocks bring significant improvements in performance for existing state-of-the-art CNNs at slight additional computational cost. Squeeze-and-Excitation Networks formed the foundation of our ILSVRC 2017 classification submission which won first place and reduced the top-5 error to 2.251 percent, surpassing the winning entry of 2016 by a relative improvement of ${\sim }$ ∼ 25 percent. Models and code are available at https://github.com/hujie-frank/SENet .

Pattern Recognition and Machine Learning

Abstract: Probability Distributions.- Linear Models for Regression.- Linear Models for Classification.- Neural Networks.- Kernel Methods.- Sparse Kernel Machines.- Graphical Models.- Mixture Models and EM.- Approximate Inference.- Sampling Methods.- Continuous Latent Variables.- Sequential Data.- Combining Models.

CBAM: Convolutional Block Attention Module

Abstract: We propose Convolutional Block Attention Module (CBAM), a simple yet effective attention module for feed-forward convolutional neural networks. Given an intermediate feature map, our module sequentially infers attention maps along two separate dimensions, channel and spatial, then the attention maps are multiplied to the input feature map for adaptive feature refinement. Because CBAM is a lightweight and general module, it can be integrated into any CNN architectures seamlessly with negligible overheads and is end-to-end trainable along with base CNNs. We validate our CBAM through extensive experiments on ImageNet-1K, MS~COCO detection, and VOC~2007 detection datasets. Our experiments show consistent improvements in classification and detection performances with various models, demonstrating the wide applicability of CBAM. The code and models will be publicly available.

YOLOv4: Optimal Speed and Accuracy of Object Detection

Abstract: There are a huge number of features which are said to improve Convolutional Neural Network (CNN) accuracy. Practical testing of combinations of such features on large datasets, and theoretical justification of the result, is required. Some features operate on certain models exclusively and for certain problems exclusively, or only for small-scale datasets; while some features, such as batch-normalization and residual-connections, are applicable to the majority of models, tasks, and datasets. We assume that such universal features include Weighted-Residual-Connections (WRC), Cross-Stage-Partial-connections (CSP), Cross mini-Batch Normalization (CmBN), Self-adversarial-training (SAT) and Mish-activation. We use new features: WRC, CSP, CmBN, SAT, Mish activation, Mosaic data augmentation, CmBN, DropBlock regularization, and CIoU loss, and combine some of them to achieve state-of-the-art results: 43.5% AP (65.7% AP50) for the MS COCO dataset at a realtime speed of ~65 FPS on Tesla V100. Source code is at this https URL

Squeeze-and-Excitation Networks

Abstract: The central building block of convolutional neural networks (CNNs) is the convolution operator, which enables networks to construct informative features by fusing both spatial and channel-wise information within local receptive fields at each layer. A broad range of prior research has investigated the spatial component of this relationship, seeking to strengthen the representational power of a CNN by enhancing the quality of spatial encodings throughout its feature hierarchy. In this work, we focus instead on the channel relationship and propose a novel architectural unit, which we term the "Squeeze-and-Excitation" (SE) block, that adaptively recalibrates channel-wise feature responses by explicitly modelling interdependencies between channels. We show that these blocks can be stacked together to form SENet architectures that generalise extremely effectively across different datasets. We further demonstrate that SE blocks bring significant improvements in performance for existing state-of-the-art CNNs at slight additional computational cost. Squeeze-and-Excitation Networks formed the foundation of our ILSVRC 2017 classification submission which won first place and reduced the top-5 error to 2.251%, surpassing the winning entry of 2016 by a relative improvement of ~25%. Models and code are available at this https URL.