weed-detection

Weed Detection

Description

Weed identification and site-specific weed management skill for precision agriculture, supporting detection, classification, and localization of weed species in field conditions.

When to Use

• User needs to detect or identify weed species in agricultural fields
• Need to build precision herbicide application systems
• Need to compare weed detection methods across different field conditions
• Need to work with DeepWeeds or similar weed datasets

Prerequisites

pip install torch torchvision ultralytics timm albumentations

Supported Datasets

Dataset	Classes	Images	Environment
DeepWeeds	9 weed species	17,509	Australian field

Quick Start

from agri_cv_research.models import YOLODetector, ViTClassifier
from agri_cv_research.datasets import DeepWeedsDataset

dataset = DeepWeedsDataset(root="./data/deepweeds", split="train", download=True)

model = YOLODetector(model_size="n", task="detect", num_classes=9)
model.train(dataset, epochs=50, batch_size=32)
metrics = model.evaluate(dataset, split="test")

Code Examples

Multi-Model Weed Classification

from agri_cv_research.models import ViTClassifier

weeds_dataset = DeepWeedsDataset(split="train", download=True)

classifiers = {
    "ResNet-50": ViTClassifier(backbone="resnet50"),
    "EfficientNet-B4": ViTClassifier(backbone="efficientnet_b4"),
    "ViT-B/16": ViTClassifier(backbone="vit_base_patch16_224"),
}

for name, clf in classifiers.items():
    clf.train(weeds_dataset, epochs=30)
    results[name] = clf.evaluate(weeds_dataset, split="test")