Oryza Mind — Rice Disease Segmentation Model
This is an Academic project for Rice Paddy Farmers- This repository hosts the vision model that powers Oryza Mind, an agent that detects and segments rice plant diseases from photos and feeds the result to a Gemini-based reasoning layer for treatment recommendations.
Model
- Architecture: YOLOv11 (segmentation)
- Task: Instance segmentation of rice disease lesions/regions on leaf images
- Framework: Built with Ultralytics YOLO
- Formats provided:
oryza-mind-seg.pt— original Ultralytics checkpoint. Use this if you want to fine-tune, retrain, or continue training on new data.oryza-mind-seg.onnx— exported for inference. Use this if you just want to run predictions, in any language/runtime with ONNX Runtime.
Training data
this model owas trained on a combined dataset made up from these two:
Dataset attribution
Trained on a combined rice-leaf-disease instance segmentation dataset. One source, published under CC BY 4.0:
- Rice-Leaf-Disease Dataset, Roboflow Universe, 2026. https://universe.roboflow.com/landebeau7/rice-leaf-disease-s1asn-nzefl (This project was originally forked from another Roboflow Universe dataset; the upstream source is no longer traceable from the public page after Roboflow's recent site changes.)
this other dataset : https://app.roboflow.com/landebeau7/rice-leaf-segmentation-cgcgd
Usage
ONNX Runtime (inference only)
```python import onnxruntime as ort
session = ort.InferenceSession("oryza-mind-seg.onnx")
preprocess your image to the model's expected input shape, then:
outputs = session.run(None, {"images": input_tensor}) ```
Ultralytics (.pt — inference or further training)
```python from ultralytics import YOLO
model = YOLO("oryza-mind-seg.pt") results = model.predict("leaf.jpg")
to keep training:
model.train(data="your_dataset.yaml", epochs=50) ```
License
This model is released under AGPL-3.0, in line with the license of the Ultralytics YOLO framework it was built on and trained with. Full credit and thanks to Ultralytics for the open YOLO architecture and training tooling that made this model possible.
Related
- Project repo:
https://github.com/BeauBryanDev/oryza_mind - Live agent:
oryza.tensorgeek.com
- Downloads last month
- -