Project notes
Foveated neural video compression
Exploring how gaze-adaptive sampling and learned video codecs can work together for streamed VR.
- Discipline
- VR & XR
- Context
- Research in progress · Unpublished
- Status
- Ongoing
Example source image from a VR video.
3 images · Select a thumbnail to explore. Original project photographs and illustrations.
The work
My ongoing research explores foveated neural video compression for VR, with an interest in visual quality and efficient streaming.
Tools & focus
Neural video compressionPyTorchPythonC++GLSLCUDAPerceptual
evaluation
Research status
Research in progress; no paper published yet. Current evidence comes from offline GPU experiments on sampled VR video.