Explore extreme low-light imaging techniques in this 31-minute lecture from the University of Central Florida. Delve into the challenges of capturing images in near-dark conditions and learn about innovative solutions using deep learning approaches. Examine the See-in-the-Dark Dataset, camera setups, and amplification factors. Understand the intricacies of Bayer arrays and X-Trans sensors. Study the network architecture and training process of a ConvNet Block UNet model. Analyze experimental results comparing traditional methods, denoising techniques like BM3D, and smartphone capabilities. Gain insights into evaluation metrics and quantitative outcomes for improving low-light photography and imaging technologies.
Learning to See in the Dark - Extreme Low Light Imaging Techniques
University of Central Florida via YouTube
Overview
Syllabus
Problem: Extreme Low Light Imaging
Related work
ISO (International Standards Organization)
Other Solutions Can we change the camera configuration?
Problem Description Using deep learning data driven approach to achieve low light imaging
Contributions
See-in-the-Dark Dataset
Camera Setup - Output
Amplification Factor (y)
Bayer Array
X-Trans vs. Bayer
Network Architecture - Training
ConvNet Block - UNet
Experimental Evaluation
Experiment 1a - Perceptual
Baseline Methods - Denoising
Experiment 1b: Qualitative - Traditional
Experiment 1b : Qualitative - BM3D
Experiment 2: Qualitative - Smartphones
Evaluation Metrics
Quantitative Results
Conclusion
Ending Remarks
Taught by
UCF CRCV
