Weed-Detection: Real-Time Serverless Weed Detection

Requirements

Please use Yolov8 on Python 3.11 and not Python 3.12. At the time of writing, TFjs has issues in Python 3.12 which prevents successful conversion of the model to TF.js

$ pip install ultralytics

Modeling

The data for model training was combined from two Kaggle datasets and can be found here

The modeling steps were performed on Pycharm, while the web deployment was performed on Visual Studio Code.

1. Training

Run this code to train the model based on the pretrained weights yolov8n from COCO.

$ yolo detect train data=coco8.yaml model=path/to/yaml epochs=100 imgsz=640

To train the on a custom dataset, on your custom data file, replace path/to/yaml with the path from you current directory to the yaml with your data file.

After training, you can get the model weights from runs/detect/your most recent run/best.pt

To export the model to TFjs format, make sure tensorflow and all its depedencies are installed and run

$ yolo export model=path/to/best.pt format=tfjs

The model will be ran on

2. Inference

To test the models performance, you can run

$ yolo detect val model=path/to/best.pt

To test it on the testing dataset, change the validation path in the yaml to your testing path

Deployment

Clone this repository

Copy nameofyourwebmodel to ./public

Update modelName in App.jsx to new model name

...
// model configs
const modelName = "nameofyourwebmodel"; // change to new model name
...

Scripts for testing/building

Make sure you change the base parameter in the vite.config.js file to the name of your github repository or else the source page will look for the wrong file

yarn start # Start dev server
yarn build # Build for productions

Acknowledgement

The starting model used is modified from the Ultralytics model. Thanks to Hyuto, for his creation of the initial javascript implementation.