Working Smart - Machine Learning in Art Asset Creation for the Games Industry
summary
Claims a production pipeline that turns real objects into game-ready assets using CT-based digital twins (not photogrammetry), with trainable segmentation, Otsu thresholding, clustering and reconstruction to keep vertex quality. Examples include bulbs, pens, phones and sushi; a maki roll goes from scan to textured STL in 15 minutes to an hour. Targets automated, physics-driven object reconstruction by 2021 and is ingesting ~750 objects for training; deployment platform exists but remains proprietary.
AI-written summary · about
chapters (9)
- 0:00Intro
- 2:01What is Machine Learning? (And why should the games industry care about it?)
- 5:28Image segmentation for Art in Gan
- 7:39Practical Use Cases
- 10:08Trainable segmentation For objects and motion
- 12:27Clustering Data and Threshold Techniques
- 16:04Clustering for Motion Graphics & Moving Ima
- 17:59How We Use ML in CT
- 22:16Helpful Things
related talks
01h 02mart tools: A new era of creativity
57mAImages - how neural networks will affect the production of games
31mTurning the real world into digital assets
56mScans in Game Production
34mGenerative AI + Spatial Computing = ∞
47m3D Digital Avatars with Machine Learning
27mAI for social good
49mState of the Art: The Future of Visual Development.