narumiruna / efficientnet-pytorch Goto Github PK

View Code? Open in Web Editor NEW

109.0 2.0 31.0 353 KB

A PyTorch implementation of "EfficientNet: Rethinking Model Scaling for Convolutional Neural Networks".

License: MIT License

Python 99.34% Dockerfile 0.25% Shell 0.42%

pytorch efficientnet imagenet mnist cifar10

efficientnet-pytorch's Introduction

EfficientNet

https://arxiv.org/abs/1905.11946

Prerequisites

Ubuntu
Python 3
- torch 1.0.1
- torchvision 0.2.2.post3
- tqdm
- mlconfig

Usage

Torch Hub

model = torch.hub.load('narumiruna/efficientnet-pytorch', 'efficientnet_b0', pretrained=True)

Train

$ python train.py -c /path/to/config

Evaluate

$ python evaluate.py --arch efficientnet_b0 -r /path/to/dataset

Pretrained models

Source: https://github.com/tensorflow/tpu/tree/master/models/official/efficientnet

Model Name	Top-1 Accuracy
efficientnet_b0	76.52%
efficientnet_b1	77.80%
efficientnet_b2	78.83%
efficientnet_b3	80.19%
efficientnet_b4	82.27%
efficientnet_b5	83.11%
efficientnet_b6
efficientnet_b7

References

efficientnet-pytorch's People

Contributors

Stargazers

Watchers

efficientnet-pytorch's Issues

imagenet already exists.

Good morning!!

There was a problem trying to use your code.

I already have an imagenet, but I don't know how to specify the path in that code.

Any help would be greatly appreciated!!!!

How to compute top-1 accuracy with this code?

Hi,
Thank you for such a nice repository.
Just a question: How to compute top-1 accuracy using this code?

Hyperparams for ImageNet

Hi, have you found a good set of hyperparameters for Imagenet?

How can I do transfer learning with my image dataset

Hi, @narumiruna

Have a good look at the code.

But how can I do transfer learning with my image dataset, not imagenet, cifar10, mnist?

I wonder..

Thanks in advance

Best,
@bemoregt.

Run inference on the model

How may I setup thi repo to run and test the inference on the models b0,b1,b2 etc?
I's appreciate your cooperation

What are the network layers of efficientnetB0-B7?

Today I tried to use this implementation on custom dataset, as you said in readme file for training on custom dataset we have to add path to training dataset by -r flag , but in the train.py there is no such flag? So could you please help me about how to train efficientnet-b0 on custom dataset?

Does this repo achieve SOTA accuracy after training for 300 epochs?

Hi
I would like to know if this code really achieves ~76% top1 accuracy with B0 on Imagenet after 300 epochs.
There are multiple things lacking (smooth cross entropy, augmentations, etc.) and I don't know how it can affect the results.
Thanks for you answer

Thanks

ImportError: cannot import name 'ModelFactory'

bro i cant find your model factory. can assist?