Dense depth from game engines: dataset + Lotus-recipe pretrained and NYU-fine-tuned models.
AI & ML interests
None defined yet.
Recent Activity
Organization Card
Rights-cleared multimodal training data captured directly from game engines. Ten modalities of ground truth, frame-locked on one clock. Origin Lab builds the premium data layer for Artificial World Intelligence™.
Start here: game-depth,
48,615 licensed RGB frames with dense engine depth. A model trained only on
game frames beats purpose-built synthetic on real KITTI with 4x less data.
Checkpoints in lotus-game-depth.
Then: frame-synced-multiplayer,
six- and eight-player live matches captured on one shared frame clock across
independent machines, sub-frame verified, with per-frame cross-player
annotations.
Format reference: game-recordings-v3,
the single-session layout both releases build on.
Access: every release ships under the Origin Lab Data License with a 90-day internal evaluation track for labs. Commercial licensing at originlab.ai/hf.