HyperAI超神经

首页资讯最新论文教程数据集百科 SOTA LLM 模型天梯 GPU 天梯顶会

中文

HyperAI超神经

Video Prediction On Kinetics 600 12 Frames

评估指标

Cond

FVD

Pred

评测结果

各个模型在此基准测试上的表现结果

模型名称	Cond	FVD	Pred	Paper Title	Repository
Video Transformer	5	170±5	11	Scaling Autoregressive Video Models	-
OmniTokenizer-AR	-	32.9	-	OmniTokenizer: A Joint Image-Video Tokenizer for Visual Generation	-
LARP	5	5.1	11	LARP: Tokenizing Videos with a Learned Autoregressive Generative Prior	-
DVD-GAN-FP	5	69.15±0.78	11	Adversarial Video Generation on Complex Datasets	-
LVT	5	224.73	11	Latent Video Transformer	-
RIN (1000 steps)	-	10.8	-	Scalable Adaptive Computation for Iterative Generation	-
TriVD-GAN-FP	5	25.74±0.66	11	Transformation-based Adversarial Video Prediction on Large-Scale Data	-
MAGVIT (-L-FP)	5	9.9±0.3	11	MAGVIT: Masked Generative Video Transformer	-
MAGVIT-v2	-	4.3±0.1	-	Language Model Beats Diffusion -- Tokenizer is Key to Visual Generation	-
CCVS	5	55±1	11	CCVS: Context-aware Controllable Video Synthesis	-
MAGVIT (-B-FP)	5	24.5±0.9	11	MAGVIT: Masked Generative Video Transformer	-
W.A.L.T.-L	-	3.3	-	Photorealistic Video Generation with Diffusion Models	-
Video VQ-VAE FVD	4	64.30±2.04	12	Predicting Video with VQVAE	-
RIN (400 steps)	-	11.5	-	Scalable Adaptive Computation for Iterative Generation	-
RaMViD	5	16.46	11	Diffusion Models for Video Prediction and Infilling	-

0 of 15 row(s) selected.