Dubaozeng

BCD-for-DNNs-PyTorch

bcd_dnn

Block Coordinate Descent Methods in Deep Learning

BlockCoordinateDescent

Code for the paper "Let’s Make Block Coordinate Descent Go Fast"

Continous-Sign-Language-Translator-

Sign language interpreters are currently required for interpreting the speech impaired people. This skill-based job of interpreters is cumbersome and hence the number of interpreters per capita across majority countries are very low or decreasing. We aim to harness technology in developing a powerful continuous sign language gestures recognition system. This computer vision-based approach will be used to recognise Argentinian sign language gestures from a video. Translating these sign language gestures is considered a monumental task in this field. The project proposes to investigate whether sign language gestures can be recognised by using a trained modified Inception V3 working as a feature selector and classifier, with a LTSM Recurrent Neural Network. Two separate approaches have been applied to recognise the Argentinian gestures. The Global Max Pooling approach outperforms the SoftMax approach, with a model accuracy of 86.10% on validation set and 75.2% on test set. Using the Inception V3 model as a feature extractor for LTSM RNN worked more efficiently and produced better results than using the Inception V3 model as a classifier. These results show the effectiveness of the research conducted. This research will help in classifying and recognising continuous sign language gestures based on machine vision. This in turn will assist people that are affected by speech and hearing impairment in understanding, translating and recognising sign gestures.

CPP

small cpp project

Deep-Learning-with-TensorFlow-book

深度学习入门开源书，基于TensorFlow 2.0案例实战。Open source Deep Learning book, based on TensorFlow 2.0 framework.

lasso

implementation of paper (Accelerated Mini-batch Randomized Block Coordinate Descent Method) and linear regression method (Lasso)

Lasso-1

This is implementation of Coordinate Descent for Lasso.

lightweight-human-pose-estimation.pytorch

Fast and accurate human pose estimation in PyTorch. Contains implementation of "Real-time 2D Multi-Person Pose Estimation on CPU: Lightweight OpenPose" paper.

MLBox

Machine Learning Algorithms implementations

Dubaozeng/BCD-for-DNNs-PyTorch

Dubaozeng/bcd_dnn

Block Coordinate Descent Methods in Deep Learning

Dubaozeng/BlockCoordinateDescent

Code for the paper "Let’s Make Block Coordinate Descent Go Fast"

Dubaozeng/Continous-Sign-Language-Translator-

Sign language interpreters are currently required for interpreting the speech impaired people. This skill-based job of interpreters is cumbersome and hence the number of interpreters per capita across majority countries are very low or decreasing. We aim to harness technology in developing a powerful continuous sign language gestures recognition system. This computer vision-based approach will be used to recognise Argentinian sign language gestures from a video. Translating these sign language gestures is considered a monumental task in this field. The project proposes to investigate whether sign language gestures can be recognised by using a trained modified Inception V3 working as a feature selector and classifier, with a LTSM Recurrent Neural Network. Two separate approaches have been applied to recognise the Argentinian gestures. The Global Max Pooling approach outperforms the SoftMax approach, with a model accuracy of 86.10% on validation set and 75.2% on test set. Using the Inception V3 model as a feature extractor for LTSM RNN worked more efficiently and produced better results than using the Inception V3 model as a classifier. These results show the effectiveness of the research conducted. This research will help in classifying and recognising continuous sign language gestures based on machine vision. This in turn will assist people that are affected by speech and hearing impairment in understanding, translating and recognising sign gestures.

Dubaozeng/CPP

small cpp project

Dubaozeng/Deep-Learning-with-TensorFlow-book

深度学习入门开源书，基于TensorFlow 2.0案例实战。Open source Deep Learning book, based on TensorFlow 2.0 framework.

Dubaozeng/lasso

implementation of paper (Accelerated Mini-batch Randomized Block Coordinate Descent Method) and linear regression method (Lasso)

Dubaozeng/Lasso-1

This is implementation of Coordinate Descent for Lasso.

Dubaozeng/lightweight-human-pose-estimation.pytorch

Fast and accurate human pose estimation in PyTorch. Contains implementation of "Real-time 2D Multi-Person Pose Estimation on CPU: Lightweight OpenPose" paper.

Dubaozeng/MLBox

Machine Learning Algorithms implementations

Dubaozeng/pytorch-siamese

Siamese Network implementation using Pytorch

Dubaozeng/RLEngine

A simple reinforcement learning simulation engine for OpenAI's gym.

Dubaozeng/S2VT

Tensorflow implement of paper: Sequence to Sequence: Video to Text

Dubaozeng/slt

Sign Language Transformers (CVPR'20)

Dubaozeng/video_to_sequence

Implementation of "Sequence to Sequence – Video to Text"