Using a pre-trained efficientnet (for experimental purposes) to classifier plant diseases given an image.
The efficient net paper can be found here and the pretrained pytorch model got from torch hub
using torch.hub.list('rwightman/gen-efficientnet-pytorch')
For more cool PyTorch pre-trained models, check out Ross Wightman pytorch models repo
Download the dataset from the following link: https://www.crowdai.org/challenges/plantvillage-disease-classification-challenge
To set up your python environment to run the code in this repository, follow the instructions below.
- Create (and activate) a new environment with Python 3.6.
- Linux or Mac:
conda create --name py39 python=3.9.2 conda activate py39
alternatively, use virtual environments if you don't have Anaconda installed.
- Clone the repository (if you haven't already!), and navigate to the
angafolder. Then, install several dependencies.
git clone https://github.com/mrdvince/anga
cd angaA minimal API endpoint to expose your model, you can make it more robust if you want.
Run uvicorn api.main:app and visit you local ip if running locally port 8000 playground to access the interactive docs (included by default). i.e. http://127.0.0.1:8000/playground
See the README on this link has been(forked from). The readme contains information on the folder structures and how to modify the hyperparameters to your liking.
On a high level:
- the config json file contains the model hyperparamaters and other settings
Run python train.py -c config.json to train the model.
Logged using tensorboard
