Agri-CV-Research

Agri-CV Auto Research

Description

Automated research framework for agricultural computer vision. Input a research direction and the system automatically completes literature search, hypothesis generation, experiment execution, result analysis, and paper writing.

When to Use

• •User wants to do agricultural/crop/plant/disease related CV research
• •User wants to automate the research workflow
• •User needs to quickly generate a paper draft in the agricultural CV direction
• •User needs to compare different model performances on agricultural datasets

Quick Start

# Install
pip install -e .

# Run full pipeline
agri-research run --topic "YOUR RESEARCH TOPIC" --auto-approve

# Or use individual commands
agri-research search --topic "plant disease detection ViT"
agri-research train --config config/experiment.yaml
agri-research eval --model best.pt --dataset plantvillage
agri-research paper --results ./outputs --template computers_electronics_agri

Project Structure

• •skills/: Agent skills for agricultural CV tasks (K-Dense format)
• •pipeline/: 9-stage research pipeline
• •datasets/: Dataset loaders for agricultural CV benchmarks
• •models/: Model zoo with agricultural CV SOTA models
• •evaluation/: Comprehensive evaluation framework
• •visualization/: Publication-quality figure generation
• •templates/: LaTeX templates for major venues
• •config/: YAML configuration files
• •docs/: Tutorials and case studies

Supported Datasets

• •PlantVillage (54,305 images, 38 classes, classification)
• •PlantDoc (2,598 images, 27 classes, real-field detection)
• •DeepWeeds (17,509 images, 9 classes, weed identification)
• •MinneApple (1,000+ images, apple detection & counting)

Supported Models

• •YOLOv8 (detection, classification, segmentation)
• •ViT / EfficientNet / ResNet (classification)
• •SAM (segmentation)
• •CLIP (few-shot / zero-shot)

Key Commands