turandai/4DGV_eval

This repo holds the evaluation results of the paper "4D Gaussian Videos with Motion Layering", including for N3DV, MeetRoom, Technicolor, and DeskGames datasets.

Each .npz contains {'renders': fp32 renders, 'gts': fp32 gts, 'metrics': per-frame metrics}.

Feel free to use for comparison experiments, and please kindly cite us:

@article{dai2025_4dgv,
  title={4D Gaussian Videos with Motion Layering},
  author={Dai, Pinxuan and Zhang, Peiquan and Dong, Zheng and Xu, Ke and Peng, Yifan and Ding, Dandan and Shen, Yujun and Yang, Yin and Liu, Xinguo and Lau, Rynson WH and Xu, Weiwei},
  journal={ACM Transactions on Graphics (TOG)},
  volume={44},
  number={4},
  year={2025},
  doi={10.1145/3731189},
  publisher={ACM New York, NY, USA}
}

百度网盘：https://pan.baidu.com/s/1fprYkNOvUK_32e8IXVCY4g ， 提取码: 4dgv

Per-scene summary:

Dataset	Scene	PSNR	SSIM	LPIPS	Size (MB)
N3DV	Coffee Martini	28.814	0.9195	0.1447	19.05
	Cook Spinach	34.008	0.9582	0.1250	24.87
	Cut Beef	33.948	0.9578	0.1306	24.46
	Flame Salmon	29.491	0.9255	0.1399	17.26
	Flame Steak	34.332	0.9648	0.1148	23.94
	Sear Steak	34.730	0.9668	0.1143	17.69
	Mean	32.5538	0.9488	0.1282	21.2117
MeetRoom	Discussion	32.649	0.9631	0.1595	31.35
	Step In	30.944	0.9497	0.1816	13.24
	Trimming	33.545	0.9596	0.1698	16.64
	VR Headset	32.106	0.9551	0.1806	15.61
	Mean	32.3110	0.9569	0.1729	19.2100
Technicolor	Birthday	29.198	0.9095	0.0589	87.92
	Fabien	32.984	0.8523	0.3262	17.68
	Painter	35.004	0.9183	0.2072	34.30
	Theater	31.532	0.8673	0.2686	41.13
	Train	30.108	0.9193	0.1241	20.15
	Mean	31.7652	0.8933	0.1970	40.2360
DeskGames	Cube	30.099	0.9089	0.1828	25.43
	Domino	31.456	0.9233	0.1663	35.53
	Reading	31.659	0.9258	0.1673	29.16
	UNO	30.357	0.9232	0.1550	47.78
	Mean	30.8927	0.9203	0.1678	34.4750