A CV toolkit for my papers.

How to use Cross Replica / Synchronized Batchnorm in Pytorch

A Re-implementation of Fixed-update Initialization

MXNet Gluon Synchronized Batch Normalization Preview

Rethinking the Smaller-Norm-Less-Informative Assumption in Channel Pruning of Convolution Layers https://arxiv.org/abs/1802.00124

My solutions for Assignments of CS231n: Convolutional Neural Networks for Visual Recognition

Deep learning projects including applications (face recognition, neural style transfer, autonomous driving, sign language reading, music generation, translation, speech recognition and NLP) and theories (CNNs, RNNs, LSTM, Adam, Dropout, BatchNorm, Xavier/He initialization, hyperparameter tuning, regularization, optimization, Residual Networks). …

Review materials for the TWiML Study Group. Contains annotated versions of the original Jupyter noteboooks (look for names like *_jcat.ipynb ), slide decks from weekly Zoom meetups, etc.

Stochastic Downsampling for Cost-Adjustable Inference and Improved Regularization in Convolutional Networks

Synchronized BatchNorm in PyTorch 1.0

VGG16 architecture with BatchNorm

Cross-platform mobile Neural network C library for training and inference on the device. CPU only. It fits for time-series data.

Playground repository to highlight the problem of BatchNorm layers for an blog article

MNIST Classification using Neural Network and Back Propagation. Written in Python and depends only on Numpy

remove batchnorm when deploy mxnet model

Code to fold batch norm layer of a DNN model in pytorch

Partial transfusion: on the expressive influence of trainable batch norm parameters for transfer learning. TL;DR: Fine-tuning only the batch norm affine parameters leads to similar performance as to fine-tuning all of the model parameters

Solutions for Andrej Karpathy's "Neural Networks: Zero to Hero" course

fuse batch normalization layer and con layer in CNN