Create 10 000 synthetic head-pose images in NVIDIA Omniverse Isaac Sim 4.5 -- 3
Budget: $250 – $750 NZD
I need a 3D-graphics specialist to generate approximately 10,000 photorealistic training images in NVIDIA Omniverse Isaac Sim 4.5. A single virtual camera should orbit a supplied OBJ head model, capturing varied angles and lighting. I will supply the head model obj. Note: you will need access to a 4090 GPU or better for this project.
Key requirements:
- Five furnished home rooms (living room, kitchen, bedroom, study, hallway) to maximize background diversity. Use default rooms in NVIDIA Omniverse Isaac Sim.
- Outputs per frame:
- RGB PNG
- Depth EXR
- Binary head mask PNG
- JSON with camera-to-head 6-DoF transform
- Accurate camera settings (to be supplied)
- Lighting and pose randomization; total 10k frames split roughly 2k per room.
- Deliver USD scenes, render script, and README so I can rerun the pipeline.
- Timeline: scene block-out → prototype → full render → QA
This dataset will train a CNN for real-time head tracking in an hair-clipper. (see real footage → https://youtu.be/Jv9XeWMMqMM).
Key requirements:
- Five furnished home rooms (living room, kitchen, bedroom, study, hallway) to maximize background diversity. Use default rooms in NVIDIA Omniverse Isaac Sim.
- Outputs per frame:
- RGB PNG
- Depth EXR
- Binary head mask PNG
- JSON with camera-to-head 6-DoF transform
- Accurate camera settings (to be supplied)
- Lighting and pose randomization; total 10k frames split roughly 2k per room.
- Deliver USD scenes, render script, and README so I can rerun the pipeline.
- Timeline: scene block-out → prototype → full render → QA
This dataset will train a CNN for real-time head tracking in an hair-clipper. (see real footage → https://youtu.be/Jv9XeWMMqMM).