avinashreddy09/solar-defect-models

Solar Defect Models

Trained models for drone-based solar panel defect detection.

Models

File	Purpose	Format	Size
micro_crack_el_detector.onnx	EL micro-crack detection	ONNX	84.4 MB
thermal_yolov8n_best.pt	Thermal hotspot detection	PyTorch	6.26 MB

Usage

from huggingface_hub import hf_hub_download
from ultralytics import YOLO

# Micro-crack detector (EL images)
path = hf_hub_download('avinashreddy09/solar-defect-models', 'micro_crack_el_detector.onnx')
model = YOLO(path)

# Thermal detector (thermal IR images)
path = hf_hub_download('avinashreddy09/solar-defect-models', 'thermal_yolov8n_best.pt')
model = YOLO(path)

Classes

Thermal (8): Single_Hotspot, Multi_Hotspot, Single_Diode, Multi_Diode, Single_Bypassed_Substring, Multi_Bypassed_Substring, String_Open_Circuit, String_Reversed_Polarity

Micro-crack (7): black_core, crack, finger, horizontal_dislocation, short_circuit, star_crack, thick_line

Training Data

• Thermal: ThermoSolar-PV (Kaggle: pkdarabi/solarpanel)
• Micro-crack: Pre-trained from azizasyd/yolov11-el-fault-detector

License

MIT