|
Download README.md from Yusin/ReTokenGS-stream: direct link, hf CLI and curl.
- Browser
- Download file 2.19 kB
-
https://huggingface.co/Yusin/ReTokenGS-stream/resolve/main/README.md
- Command line
-
hf download hf://Yusin/ReTokenGS-stream/README.md
-
curl -L -o README.md https://huggingface.co/Yusin/ReTokenGS-stream/resolve/main/README.md
2.19 kB
metadata
license: apache-2.0
tags:
- gaussian-splatting
- 3d-reconstruction
- novel-view-synthesis
- feed-forward
- streaming
datasets:
- DL3DV/DL3DV-10K
ReTokenGS β stream
Recurrent Gaussian Tokens with uncertainty-driven births and streaming input views, built on LocusGS / TokenGS.
- The loop. One weight-shared decoder layer runs as a loop. Every round, each token reads the rendering gradient of its own Gaussians (G3R) and takes a line-searched latent gradient step.
- Births. The tokens whose learned uncertainty head predicts the largest error split, and the number of splits per round is the predicted error itself. The loop starts at 256 tokens and grows to at most 1024, with 32 Gaussians per token.
- Streaming. Input views can join at any round. An arriving view is encoded on arrival, and all tokens keep iterating on every view seen so far.
The model works with posed views at 256Γ256. model.safetensors is epoch 29 of the streaming fine-tune.
| Official DL3DV benchmark (140 scenes, 4 views, 256Γ256, 32 rounds) | PSNR | LPIPS | SSIM |
|---|---|---|---|
ReTokenGS-grad (Yusin/ReTokenGS-grad) |
24.733 | 0.2664 | 0.7916 |
ReTokenGS-dens, 4096 tokens (Yusin/ReTokenGS-dens) |
25.091 | 0.2442 | 0.8031 |
| ReTokenGS-stream | 22.856 | 0.3881 | 0.7108 |
| Streaming (same 140 scenes, 6-view split, PSNR) | 16 rounds | 24 rounds | 32 rounds |
|---|---|---|---|
| all 6 views from round 1 | 23.421 | β | 24.074 |
| views 1β4 only | β | 22.817 | β |
| views 1β4, views 5β6 join at round 17 | β | 23.289 | 23.675 |
In the standard 4-view setting this model is behind grad and dens. Its contribution is the birth rule and the streaming mechanism: the late views add +0.47 dB over the views-1β4 reference at the same number of rounds.
Code, evaluation and training commands: GitHub Yusin2Chen/TokenGS-Loop, branch stream. License: Apache-2.0 (derived from LocusGS / TokenGS).