Project description

Goal

This is the GitHub repository by Group 14. In this project, we employ an image classification model to classify images of crop and weed seedlings. The goal is to address a challenge in modern agriculture: improving efficiency and reducing labor while meeting increasing food demands.

Humans have been practicing agriculture for over 12,000 years and have continuously worked to increase both yield and efficiency. Over time, advancements in technology, methods, and tools have transformed the way we cultivate crops. However, one major challenge in agriculture today is meeting the growing global demand for food while grappling with limited resources, such as arable land, water, and labor. Weeds present a significant obstacle, as they compete with crops for essential nutrients and water, potentially reducing yields dramatically.

Weed removal is labor-intensive and often relies heavily on the use of herbicides, which can have long-term environmental and economic costs. The purpose of this project is to leverage machine learning to develop a model that can accurately distinguish between weed seedlings and crop seedlings. This model could serve as the foundation for autonomous systems capable of identifying and removing weeds efficiently. Such systems have the potential to revolutionize agriculture by reducing manual labor, minimizing herbicide use, and ultimately increasing crop yield in a sustainable manner.

Framework

In this project, we utilize the TIMM (PyTorch Image Models) framework to access and implement image classification architectures. TIMM provides a wide range of pre-trained models, enabling us to select and fine-tune a suitable architecture for our specific classification task.

Data

The data we plan to use is found at

https://www.kaggle.com/datasets/vbookshelf/v2-plant-seedlings-dataset

which has 12 classes with 5539 images. These classes consist of different common plant species in Denmark. In this dataset the classes have some imbalance. The data is structured in one folder for each class, which should all be placed in data/raw such that the structure becomes:

├── data/                     # Data directory
│   ├── processed
│   └── raw
│       ├── Black-grass
│       ├── Charlock
        ...

To process data, execute:

python src/plant_seedlings/data.py

Environment setup

To run the project, start by creating a .env file containing the folloring arguments:

WANDB_API_KEY=<your api key>
WANDB_PROJECT=<your project name>

To finish project setup, execute:

invoke requirements

Model

In the project we make use of two pretrained models and one custom model:

resnet: Pretrained ResNet_18 TIMM model

mobilenet: Pretrained mobilenetv3_small_050 TIMM model

custom: Basic nn.Module Model

Included in the project is one set of weights for each model trained on the data set. Additional training is setup using config.yaml and training_conf.yaml. To start training, execute:

python src/plant_seedlings/train.py

Project structure

The initial directory structure of the project is created using mlops_template, a cookiecutter template with the following structure:

├── .github/                  # Github actions and dependabot
│   ├── dependabot.yaml
│   └── workflows/
│       └── tests.yaml
├── configs/                  # Configuration files
├── data/                     # Data directory
│   ├── processed
│   └── raw
├── dockerfiles/              # Dockerfiles
│   ├── api.Dockerfile
│   └── train.Dockerfile
├── docs/                     # Documentation
│   ├── mkdocs.yml
│   └── source/
│       └── index.md
├── models/                   # Trained models
├── notebooks/                # Jupyter notebooks
├── reports/                  # Reports
│   └── figures/
├── src/                      # Source code
│   ├── project_name/
│   │   ├── __init__.py
│   │   ├── api.py
│   │   ├── data.py
│   │   ├── evaluate.py
│   │   ├── models.py
│   │   ├── train.py
│   │   └── visualize.py
└── tests/                    # Tests
│   ├── __init__.py
│   ├── test_api.py
│   ├── test_data.py
│   └── test_model.py
├── .gitignore
├── .pre-commit-config.yaml
├── LICENSE
├── pyproject.toml            # Python project file
├── README.md                 # Project README
├── requirements.txt          # Project requirements
├── requirements_dev.txt      # Development requirements
└── tasks.py                  # Project tasks