Get the latest tech news
Watermark segmentation
Open-source watermark segmentation by DiffusionDynamics.ai and clear.photo. Harness deep learning plus synthetic data augmentation in PyTorch to detect logos/text precisely. This minimal code refer...
The goal is to provide a clear, understandable baseline that is easy to modify and build upon, even allowing for fine-tuning on consumer hardware like laptops with Apple M-series chips. It dynamically applies logo watermarks (from the logos/ directory) onto background images with randomized properties (scale, rotation, opacity, position, blend mode) to create diverse training samples and their corresponding ground truth masks. This involves synthetic data augmentation: dynamically applying diverse and randomized watermarks (varying size, position, opacity, rotation, blend modes) onto clean background images during training.
Or read this on Hacker News