Oriel — AI & technology worth noticing, every day.

speridlabs/idol

AI model · https://huggingface.co/speridlabs/idol

First seen by Oriel on .

AI summary

Idol converts a single object image into a 3D Gaussian splat with up to 1,048,576 Gaussians, including unseen sides, and is designed to run on consumer GPUs.

  • It removes the background, frames the object, predicts the input camera, and exports a standard 3DGS .ply file.
  • The package supports Python 3.10+, PyTorch 2.5+, and NVIDIA CUDA GPUs; its weights and gated image encoders download automatically on first use.
  • Peak VRAM use is reported as 7.2 GB, with roughly 3.7–4.4 GB occupied by model weights.
  • GPU sampling optimization is enabled by default and is reported to be about 1.7× faster without changing output quality; the first image takes longer because compilation runs.
  • In the 1 × H100 benchmark, Idol took 6.8 seconds and led several image-quality metrics, while LGM had the highest PSNR but visibly blurry results.

#3d gaussian splatting #3d reconstruction #computer vision #gpu inference

Generated by gpt-5.6-sol. Check the original for details.

Metric trends

Last 30 days · Up to 120 recorded samples per source and metric, with reposts kept separate · First and latest retained

Likes

Total likes reported by the source.

  • Hugging Face Models: 2,433, latest observation ; +2,330 over observed period since

Downloads (30d)

Rolling 30-day downloads. A change in this value is not new downloads during the observed period.

  • Hugging Face Models: 34,752, latest observation ; +34,531 over observed period since

Aliases

SourceExternal IDURL
Hugging Face Modelssperidlabs/idolhttps://huggingface.co/speridlabs/idol Read discussion