Skip to content

InahL/BeeVision

Repository files navigation

BeeVision

BeeVision is an AI-powered pipeline for detecting, tracking, and classifying bees and wasps. By combining object detection, pose estimation, species classification, and model explainability, the system enables non-invasive monitoring of insect behavior to support ecological conservation and hive health analysis.

Developed as part of the AI for Conservation (AI4C) project


Conservation Motivation

Bees pollinate over 70% of global crops, yet their populations are under severe threat due to pesticides and climate change. Wasps, while ecologically important as predators, also pose risks to beehives. Accurately distinguishing bees from wasps and understanding their movement is crucial for developing targeted and minimally invasive conservation strategies.


Key Features

Module Description
Detection YOLOv8-based bee detection with bounding box outputs
Pose Estimation YOLOv8-pose for head/tail orientation of bees
Classification ResNet50-based bee vs wasp classification
Explainability Grad-CAM visualizations to understand model focus
Pipeline Modular processing with Snakemake automation

All outputs are stored in CSV format, supporting further analysis or integration into hardware systems.


Project Structure

BeeVision/
├── classification/        # ResNet-based bee/wasp classifier
├── detection/             # YOLOv8-based object detection
├── kaggle_bee_vs_wasp/    # Dataset: bee/wasp images and labels
├── notebooks/             # Experiment and debugging notebooks
├── pose_estimation/       # YOLOv8-pose for head/tail keypoints
├── tools/                 # Utility scripts (e.g., label generator, data sorting)
├── visualization/         # Grad-CAM visualization
├── pipeline.png           # End-to-end pipeline diagram
└── README.md              # You're here!

Technical Stack

Category Tools / Libraries
Language Python
AI Frameworks PyTorch
CV & Model Libraries OpenCV, YOLOv8-pose, Grad-CAM, Torchvision
Data Processing & Visualization Pandas, NumPy, CSV, JSON, Matplotlib, Seaborn
Workflow Automation Snakemake
Development Tools Jupyter Notebook, VS Code, GitHub

Dataset

We used open-access datasets from multiple sources:

The dataset is located in kaggle_bee_vs_wasp/.

About

No description, website, or topics provided.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Contributors 3

  •  
  •  
  •