Skip to content

Latest commit

 

History

History
27 lines (24 loc) · 1.29 KB

README.md

File metadata and controls

27 lines (24 loc) · 1.29 KB

YOLO_filament : A YOLOv5 based detection of solar chromospheric filaments (Gosain,S. 2023)

Customized YOLOv8 neural network model for detecting filaments, prominences and active regions in fulldisk H-alpha images from GONG network. YOLO (You Only Look Once) is a convolutional neural network (CNN) designed for object recognition in images and videos. It is very fast and allows real time object detection.

Install YOLOv8 Follow instructions on YOLO website https://github.com/ultralytics/ultralytics

Create Labeled Data for Training and Validation Used LabelImg to label images.

pip install labelImg

Train Model for Filaments, prominences and Active Regions using Labeled data

yolo task=detect mode=train epochs=400 data=data_custom.yaml model=yolov8m.pt imgsz=640 batch=32 patience=400

Predict using YOLOv8 on images/movies from NSO GONG Halpha network.

yolo task=detect mode=predict model=custom_halpha.pt show=True conf=0.5 source=1.jpeg

alt text

Utility programs convert_gif_mp4.py : Convert a colored or BW GIF animation to BW mp4 file convert_gif_HE_mp4.py: Convert a colored or BW GIF animation to BW mp4 file with HistogramEqualization download.py: Download images from internet search.