How shu-hdrqad/hdr-iqa-dataset works
Overall Structure and Main Components
• Data Sources:
• Dataset: Indicate the HDRImage folder containing sample HDR images (.exr files) and the Images folder that includes auxiliary analysis graphics.
• Pre-trained Model: Represent the pretrained folder where the pre-trained network is stored.
• Input Module:
• Demo Script: Start with demo.py as the entry point where an HDR image is read.
• Utility Subsystem: Show that demo.py uses utility functions (from utils/, e.g., image_io.py, exrio.py, and others) for reading images and performing necessary pre-processing.
• Core Network (Model):
• Model Module: Represent model/HDRIQA_model.py as implementing the core network.
• DRL Module: Highlight the branch that analyzes local distortions based on dynamic range and HDR artifacts.
• IRQI Module: Indicate the branch that accounts for inter-regional quality dependencies.
• Processing Flow:
• Output:
• Quality Prediction: Include a block (output layer) that represents the numerical quality score or other metrics, achieved after the model processes the image.
• From the dataset (HDR images) to the input module (demo.py).
• From demo.py to utility functions (data pre-processing).
• From pre-processed data into the network model.
• Another style for output components.