EPO v1.2 speeds pose refinement 2.4x to 318 it/s on RTX 4090
ducha_aiki · x · 2026-09-08
EPO v1.2, showcased at ECCV, makes its pose-refinement loop 2.4x faster than v1.1 (133 → 318 it/s on an RTX 4090) and 8.4x faster than the PyTorch backend, with accuracy intact: dataset means within ±0.5 AUC. Refining 150 cameras now takes 5 seconds.
More from Research
- Causal foundation models estimate causal effects in-context, no fine-tuning needed — Layer6 · 2026-09-09
- Conformal Relevance framework automates conformal score design via in-context ensembles — Layer6 · 2026-09-09
- Omnii, a language model pretrained on DNA, designs personalized mRNA cancer vaccines — exnx · 2026-09-09
- Huawei and EPFL Release New Depth Estimation Model with Sparse Point Cloud Completion — AntonObukhov1 · 2026-09-09
- Marigold V2 launches at SIGGRAPH Asia 2026: sharp diffusion-transformer depth estimation — AntonObukhov1 · 2026-09-09
- Magic claims 50x pretraining efficiency: matches DeepSeek V4 Pro for ~$0.5M — magicailabs · 2026-09-09