Introduction to Deep Learning Models in Unreal
summary
Uses a PyTorch style‑transfer model exported to ONNX and runs real‑time, frame‑by‑frame inference in Unreal Engine 5 via the Neural Network Inference plugin and ONNX Runtime. Pipeline captures a screenshot, uses the OpenCV helper to resize to 224x224 and normalize, runs GPU inference in C++, converts floats back to RGB, makes a texture and applies it as a post‑process each frame. Demo works with low latency but shows some blurriness; plugin and workflow remain experimental.
AI-written summary · about
chapters (19)
- 0:00Start
- 0:55Introduction to Deep Learning
- 2:40Real-Time Style Transfer
- 3:57Deep Learning Neural Networks
- 4:38Automating Game Logic
- 5:16Story Generation
- 5:54Speech Recognition
- 6:48Code Generation and the Codex Models
- 7:26Chess Board
- 7:59Ai Dungeon
- 11:34Automated Reachability Testing
- 13:31Style Transfer Models
- 17:32Image Classification
- 17:51Fast Neural Style Transfer
- 18:24Create New Models
- 22:41Pre-Processing
- 23:09Opencv Helper
- 29:42Add the Plugins
- 35:16Post-Processing
related talks
18mDeploying ML solutions with low latency in Python
01h 05mDeep Learning with PyTorch
32m..Stories About Industrializing In-Game Machine Learning Solutions
26mDeep Learning in OpenCV
27mEmbedded Deep Learning Super Resolution on GStreamer using ONNX Inference Runtime
41mLeverage Power of Machine Learning with ONNX
55mA Tour of Deep Learning With C++
33mNASA Lunar Terrain Visualization in Unreal Engine 5