Paper_CLR_CV / 01_models /03_trained_models /models_description.txt
MaryPazRB's picture
Upload 23 files
d72a22b verified
MODELS DESCRIPTION
==================
This folder contains the pre-trained models used in the paper.
You can use these weights entirely without training new models.
1. clr_YOLOV8.pt
----------------
* **Pipeline Phase:** 01_leaf_extraction
* **Model Type:** YOLOv8 (Ultralytics)
* **Purpose:** Detects and segments coffee leaves in field images.
* **Training Data:** Labeled using GroundedSAM (Autodistill).
* **Usage:** Used by `extract_leaves.py` to identify individual leaves.
2. severity_rust.pt
-------------------
* **Pipeline Phase:** 02_severity_segmentation
* **Model Type:** SAM2 (Segment Anything Model 2) - Fine-tuned Adapter
* **Purpose:** Segments "Rust" (disease) areas on extracted leaf images.
* **Usage:** Used by `inference_clr.py`.
3. deeplab_binary_best.pth
--------------------------
* **Pipeline Phase:** 02_severity_segmentation
* **Model Type:** DeepLabV3+ (ResNet50 Encoder)
* **Purpose:** Alternative model for segmenting "Rust" areas.
* **Usage:** Used by `deeplab/inference_deeplabv3.py`.
HOW TO USE
==========
To use these models, update the `CHECKPOINT_PATH` or `MODEL_PATH` variable in the corresponding python script to point to these files.
Example (inference_clr.py):
CHECKPOINT_PATH = "./03_trained_models/severity_rust.pt"