Learn to enhance low light images using Zero-DCE (Zero-Reference Deep Curve Estimation) in this 36-minute video tutorial. Explore an unsupervised learning approach that requires only dark, low light images to produce impressive results. Discover how to implement this technique using Python, Keras, and Weights & Biases, resulting in a compact 350 KB model capable of real-time image enhancement. Engage with Soumik Rakshit, the creator of the Keras implementation, as he discusses the strengths and weaknesses of Zero-DCE. Gain hands-on experience in tracking machine learning experiments with Weights & Biases and Keras, and learn to use W&B Tables for exploring model predictions. Follow along with a provided Google Colab notebook to train your own Zero-DCE model and visualize results on test datasets.
Enhance Low Light Images Using Keras, Python and Weights & Biases
Overview
Syllabus
Intro
What's gonna be in the video
Colab notebook and exploring data with W&B Tables
Hitting up Soumik
Use-cases for Zero-DCE
Looking through example model predictions
Why it works well on some images, but struggles on others?
A challenge for our amazing W&B community
How Zero-DCE works explained by Soumik
Why Soumik really enjoys Keras and TensorFlow
Keras and TensorFlow work really well with Weights & Biases
Thanks for chatting and explaining stuff Soumik
Training Zero-DCE
Monitoring training
Visualizing model predictions on the test dataset
Outro
Taught by
Weights & Biases
