arXiv subjects

Ziheng Wang

Publications and source records attributed to Ziheng Wang.

78 records · Page 5Linked to original sources

arXiv2020

SparseRT: Accelerating Unstructured Sparsity on GPUs for Deep Learning Inference

In recent years, there has been a flurry of research in deep neural network pruning and compression. Early approaches prune weights individually. However, it is difficult to take advantage of the resulting unstructured sparsity patterns on modern hardware like GPUs. As a result, pruning strategies which impose sparsity structures in the weights have become more popular. However,these structured pruning approaches typically lead to higher losses in accuracy than unstructured pruning. In this paper, we present SparseRT, a code generator that leverage unstructured sparsity to accelerate sparse linear algebra operations in deep learning inference on GPUs. For 1x1 convolutions and fully connected layers, we demonstrate geometric mean of speedups of 3.4x over the equivalent dense computation at 90% sparsity and 5.4x at 95% sparsity when evaluated on hundreds of test cases in deep learning. For sparse 3x3 convolutions, we show speedups of over 5x on use cases in ResNet-50.

cs.LG↗

arXiv2019

Accelerated CNN Training Through Gradient Approximation

Training deep convolutional neural networks such as VGG and ResNet by gradient descent is an expensive exercise requiring specialized hardware such as GPUs. Recent works have examined the possibility of approximating the gradient computation while maintaining the same convergence properties. While promising, the approximations only work on relatively small datasets such as MNIST. They also fail to achieve real wall-clock speedups due to lack of efficient GPU implementations of the proposed approximation methods. In this work, we explore three alternative methods to approximate gradients, with an efficient GPU kernel implementation for one of them. We achieve wall-clock speedup with ResNet-20 and VGG-19 on the CIFAR-10 dataset upwards of 7%, with a minimal loss in validation accuracy.

cs.CV↗

arXiv2019

Deep Learning with Convolutional Neural Network for Objective Skill Evaluation in Robot-assisted Surgery

With the advent of robot-assisted surgery, the role of data-driven approaches to integrate statistics and machine learning is growing rapidly with prominent interests in objective surgical skill assessment. However, most existing work requires translating robot motion kinematics into intermediate features or gesture segments that are expensive to extract, lack efficiency, and require significant domain-specific knowledge. We propose an analytical deep learning framework for skill assessment in surgical training. A deep convolutional neural network is implemented to map multivariate time series data of the motion kinematics to individual skill levels. We perform experiments on the public minimally invasive surgical robotic dataset, JHU-ISI Gesture and Skill Assessment Working Set (JIGSAWS). Our proposed learning model achieved a competitive accuracy of 92.5%, 95.4%, and 91.3%, in the standard training tasks: Suturing, Needle-passing, and Knot-tying, respectively. Without the need of engineered features or carefully-tuned gesture segmentation, our model can successfully decode skill information from raw motion profiles via end-to-end learning. Meanwhile, the proposed model is able to reliably interpret skills within 1-3 second window, without needing an observation of entire training trial. This study highlights the potentials of deep architectures for an proficient online skill assessment in modern surgical training.

cs.CV↗

arXiv2018

SATR-DL: Improving Surgical Skill Assessment and Task Recognition in Robot-assisted Surgery with Deep Neural Networks

Purpose: This paper focuses on an automated analysis of surgical motion profiles for objective skill assessment and task recognition in robot-assisted surgery. Existing techniques heavily rely on conventional statistic measures or shallow modelings based on hand-engineered features and gesture segmentation. Such developments require significant expert knowledge, are prone to errors, and are less efficient in online adaptive training systems. Methods: In this work, we present an efficient analytic framework with a parallel deep learning architecture, SATR-DL, to assess trainee expertise and recognize surgical training activity. Through an end-to-end learning technique, abstract information of spatial representations and temporal dynamics is jointly obtained directly from raw motion sequences. Results: By leveraging a shared high-level representation learning, the resulting model is successful in the recognition of trainee skills and surgical tasks, suturing, needle-passing, and knot-tying. Meanwhile, we explore the use of ensemble in classification at the trial level, where the SATR-DL outperforms state-of-the-art performance by achieving accuracies of 0.960 and 1.000 in skill assessment and task recognition, respectively. Conclusion: This study highlights the potential of SATR-DL to provide improvements for an efficient data-driven assessment in intelligent robotic surgery.

cs.CV↗

arXiv2017

A Generative Restricted Boltzmann Machine Based Method for High-Dimensional Motion Data Modeling

Many computer vision applications involve modeling complex spatio-temporal patterns in high-dimensional motion data. Recently, restricted Boltzmann machines (RBMs) have been widely used to capture and represent spatial patterns in a single image or temporal patterns in several time slices. To model global dynamics and local spatial interactions, we propose to theoretically extend the conventional RBMs by introducing another term in the energy function to explicitly model the local spatial interactions in the input data. A learning method is then proposed to perform efficient learning for the proposed model. We further introduce a new method for multi-class classification that can effectively estimate the infeasible partition functions of different RBMs such that RBM is treated as a generative model for classification purpose. The improved RBM model is evaluated on two computer vision applications: facial expression recognition and human action recognition. Experimental results on benchmark databases demonstrate the effectiveness of the proposed algorithm.

cs.CV↗

arXiv2017

A Hierarchical Probabilistic Model for Facial Feature Detection

Facial feature detection from facial images has attracted great attention in the field of computer vision. It is a nontrivial task since the appearance and shape of the face tend to change under different conditions. In this paper, we propose a hierarchical probabilistic model that could infer the true locations of facial features given the image measurements even if the face is with significant facial expression and pose. The hierarchical model implicitly captures the lower level shape variations of facial components using the mixture model. Furthermore, in the higher level, it also learns the joint relationship among facial components, the facial expression, and the pose information through automatic structure learning and parameter estimation of the probabilistic model. Experimental results on benchmark databases demonstrate the effectiveness of the proposed hierarchical probabilistic model.

cs.CV↗