Lily_and_Weed_Detection_YOLO12

🌱 Lily–Weed Recognition ONNX Model

This repository provides an ONNX inference model for recognizing lily crops and surrounding weeds. It is designed for use in precision agriculture settings, especially for complex terrain such as terraces, where large-scale machinery is inefficient or difficult to deploy.

🚜 Background

Due to the unique terrain and climatic conditions in regions such as Gansu, specialized agricultural equipment has substantial application potential. This model is part of a broader research effort to:

Enable real-time visual perception in agricultural robotics

Support multi-agent coordination systems

Complement large-scale agricultural machinery in terraced farmland

📦 Model Overview Item Description Format ONNX Task Binary image recognition Classes Lily, Weed Usage Direct inference, no conversion required Intended Use Field monitoring / robotics / precision agriculture

⚠️ This model is intended for research and demonstration purposes.

🧠 Dataset

The training dataset was collected internally in real agricultural environments. It includes multiple growth stages of both lilies and weeds, captured under natural light and environmental conditions.

Status: Private (not publicly released)

Content: crop & weed images across growth cycles

Labeling: human annotated

Dataset is not available due to internal collection protocol and field acquisition constraints.

🔧 Inference (ONNX Runtime)

Below is a minimal example using the CPU execution provider:

import onnxruntime as ort import numpy as np from PIL import Image

Load ONNX model

session = ort.InferenceSession("model.onnx", providers=["CPUExecutionProvider"])

Preprocess image

image = Image.open("sample.jpg").convert("RGB").resize((224, 224)) arr = np.array(image).astype(np.float32) / 255.0 arr = arr.transpose(2, 0, 1)[None, :] # NCHW

Run inference

inputs = {session.get_inputs()[0].name: arr} output = session.run(None, inputs) print(output)

Adjust input size (224×224) depending on your actual model.

📤 Output Format

The model outputs a single classification prediction:

Label 0 → Weed Label 1 → Lily

You may normalize or threshold logits depending on downstream demands.

🛠️ Integration Scenarios

Embedded perception for AI agriculture equipment

Autonomous farming robots

Multi-agent cooperative machinery

Edge inference on resource-limited devices

Real-time detection in terraced farmlands

⚙️ Recommended Environment

Python ≥ 3.8

onnxruntime ≥ 1.13

numpy ≥ 1.23

pillow ≥ 9.0

Example installation:

pip install onnxruntime numpy pillow

📄 License

Specify your license (MIT / Apache-2.0 / Proprietary / Research Only).

✨ Citation

If you use this model in research or academic work:

@misc{lily-weed-onnx, title={Lily–Weed Recognition Model (ONNX)}, year={2025}, author={Your Name / Institution}, howpublished={Hugging Face}, }

📬 Contact

For questions, demonstration requests, or collaboration:

Author: Your Name

Institution: Your Affiliation

Email: your@email

Additional info: project page / docs / paper (if any)

🔔 Notes

This model is not a general plant classifier.

Performance is optimized for lily–weed scenarios only.

Domain shift (new soil types, illumination, species) may reduce accuracy.

https://cdn-uploads.huggingface.co/production/uploads/69295f5882e5edc472a76435/L9RiyHKVNtZygysHb7WTN.mp4