--- license: mit tags: - 3d-gaussian-splatting - novel-view-synthesis - pose-free - feed-forward datasets: - re10k pipeline_tag: image-to-3d --- # EvenSplat: Compact Pose-Free Feed-Forward 3D Gaussian Splatting **Kyoosuk Hwang · Junyoung Kwon** · Seoul National University of Science and Technology · Capstone Design 2026 [Code](https://github.com/junyoung6215/EvenSplat) · [Method](https://github.com/junyoung6215/EvenSplat/blob/main/docs/METHOD.md) · [Reproduce](https://github.com/junyoung6215/EvenSplat/blob/main/docs/REPRODUCE.md) ![Ground truth, the unpruned base and EvenSplat with 75, 50 and 25% of the Gaussians](https://raw.githubusercontent.com/junyoung6215/EvenSplat/main/docs/figures/hero_qualitative.jpg) **EvenSplat** renders novel views from **two unposed images** using **25 to 75% of the per-pixel Gaussians** that [NoPoSplat](https://github.com/cvg/NoPoSplat) predicts. It keeps the same fraction of Gaussians in every 2 × 2 pixel cell, drops low-reliability Gaussians first, and is fine-tuned with that pruning in the loop. ## Checkpoints | File | Gaussians kept | #Gaussians | PSNR ↑ | SSIM ↑ | LPIPS ↓ | | --- | :-: | :-: | :-: | :-: | :-: | | [`retention_r25_re10k.ckpt`](https://huggingface.co/kyoosuk/EvenSplat/blob/main/retention_r25_re10k.ckpt) | 25% | 33K | 24.62 | 0.817 | 0.184 | | [`retention_r50_re10k.ckpt`](https://huggingface.co/kyoosuk/EvenSplat/blob/main/retention_r50_re10k.ckpt) | 50% | 66K | 25.01 | 0.829 | 0.169 | | [`retention_r75_re10k.ckpt`](https://huggingface.co/kyoosuk/EvenSplat/blob/main/retention_r75_re10k.ckpt) | 75% | 98K | 25.26 | 0.837 | 0.161 | | NoPoSplat (reported) | 100% | 131K | 25.03 | 0.838 | 0.160 | RealEstate10K full test set (5,590 evaluable scenes), 256 × 256, two context views, pose alignment on. At 50%, peak render memory drops from 167 MB to 87 MB. PSNR against the number of Gaussians for EvenSplat and comparison methods Each file holds model weights only (no optimizer state), about 2.45 GB. They are for evaluation; to fine-tune again, start from the base checkpoint [`kyoosuk/noposplat-olp-rel-re10k`](https://huggingface.co/kyoosuk/noposplat-olp-rel-re10k) (`epoch_1-step_75000.ckpt`). Every model was fine-tuned from it for 10K steps at its own ratio: 2 × 2 cells, r threshold 0.35, batch 1 × 4 accumulation, lr 2e-5. ## Usage Set up the environment and data as in the [repository README](https://github.com/junyoung6215/EvenSplat#requirements), then: ```bash hf download kyoosuk/EvenSplat retention_r50_re10k.ckpt --local-dir checkpoints python scripts/run_retention.py --config my_settings.yaml eval-grid --split full \ --checkpoint checkpoints/retention_r50_re10k.ckpt \ --baselines none --scores r_filter --ratios 0.5 --run-id ours_r50 --execute ``` Use the `--ratios` value that matches the file (0.25, 0.5 or 0.75). ## Acknowledgements Built on [NoPoSplat](https://github.com/cvg/NoPoSplat), and in turn on [pixelSplat](https://github.com/dcharatan/pixelsplat), [DUSt3R](https://github.com/naver/dust3r) and [CroCo](https://github.com/naver/croco). Released under the MIT License.