• Stars
    star
    627
  • Rank 71,654 (Top 2 %)
  • Language
    Python
  • License
    MIT License
  • Created about 5 years ago
  • Updated over 1 year ago

Reviews

There are no reviews yet. Be the first to send feedback to the community and the maintainers!

Repository Details

Unofficial PyTorch Reimplementation of RandAugment.

pytorch-randaugment

Unofficial PyTorch Reimplementation of RandAugment. Most of codes are from Fast AutoAugment.

Introduction

Models can be trained with RandAugment for the dataset of interest with no need for a separate proxy task. By only tuning two hyperparameters(N, M), you can achieve competitive performances as AutoAugments.

Install

$ pip install git+https://github.com/ildoonet/pytorch-randaugment

Usage

from torchvision.transforms import transforms
from RandAugment import RandAugment

transform_train = transforms.Compose([
    transforms.RandomCrop(32, padding=4),
    transforms.RandomHorizontalFlip(),
    transforms.ToTensor(),
    transforms.Normalize(_CIFAR_MEAN, _CIFAR_STD),
])

# Add RandAugment with N, M(hyperparameter)
transform_train.transforms.insert(0, RandAugment(N, M))

Experiment

We use same hyperparameters as the paper mentioned. We observed similar results as reported.

You can run an experiment with,

$ python RandAugment/train.py -c confs/wresnet28x10_cifar10_b256.yaml --save cifar10_wres28x10.pth

CIFAR-10 Classification

Model Paper's Result Ours
Wide-ResNet 28x10 97.3 97.4
Shake26 2x96d 98.0 98.1
Pyramid272 98.5

CIFAR-100 Classification

Model Paper's Result Ours
Wide-ResNet 28x10 83.3 83.3

SVHN Classification

Model Paper's Result Ours
Wide-ResNet 28x10 98.9 98.8

ImageNet Classification

I have experienced some difficulties while reproducing paper's result.

Issue : #9

Model Paper's Result Ours
ResNet-50 77.6 / 92.8 TODO
EfficientNet-B5 83.2 / 96.7 TODO
EfficientNet-B7 84.4 / 97.1 TODO

References

More Repositories

1

pytorch-gradual-warmup-lr

Gradually-Warmup Learning Rate Scheduler for PyTorch
Python
969
star
2

cutmix

a Ready-to-use PyTorch Extension of Unofficial CutMix Implementations with more improved performance.
Python
167
star
3

unsupervised-data-augmentation

Unofficial PyTorch Implementation of Unsupervised Data Augmentation.
Python
147
star
4

remote-dataloader

PyTorch DataLoader processed in multiple remote computation machines for heavy data processings
Python
66
star
5

data-science-bowl-2018

End-to-end one-class instance segmentation based on U-Net architecture for Data Science Bowl 2018 in Kaggle
Python
57
star
6

tf-lcnn

Tensorflow implementation for 'LCNN: Lookup-based Convolutional Neural Network'. Predict Faster using Models Trained Fast with Multi-GPUs
Python
40
star
7

kaggle-human-protein-atlas-image-classification

Kaggle 2018 @ Human Protein Atlas Image Classification
Python
34
star
8

simulated-annealing-for-tsp

This code is to solve traveling salesman problem by using simulated annealing meta heuristic.
C++
28
star
9

deep-object-detection-models

Deep Learning으둜 ν•™μŠ΅λœ Object Detection Model 에 λŒ€ν•΄ μ •λ¦¬ν•œ Archive μž„.
19
star
10

pystopwatch2

Multi Stopwatch for Python
Python
12
star
11

ai-starthon-2019

Codes used on AI Starthon 2019. 1st place in total.
Python
8
star
12

chat-ui-dashboard

Svelte
8
star
13

wedding-invitation

CSS
4
star
14

evonorm

Pytorch Implementation of EvoNorm which reproduces paper's result
Python
4
star
15

HttpReverseProxy

HTTP reverse proxy designed to facilitate secure access to HTTP services located within an internal network
Python
3
star
16

tbreader

TensorBoard Log Parser
Python
2
star