video-prediction 开源项目推荐（95 个）

leofan90 /Awesome-World-Models

一份关于世界模型定义以及使用世界模型进行通用视频生成、具身智能和自动驾驶的全面论文列表，包含论文、代码和相关网站。

chengtan9907 /OpenSTL

OpenSTL：时空预测学习的综合基准

danijar /dreamerv2

使用离散世界模型掌握 Atari

InternRobotics /Aether

[ICCV 2025 & ICCV 2025 RIWM Outstanding Paper] Aether: Geometric-Aware Unified World Modeling

JunyaoHu /common_metrics_on_video_quality

You can easily calculate FVD, PSNR, SSIM, LPIPS for evaluating the quality of generated or predicted videos.

thuml /predrnn-pytorch

Official implementation for NIPS'17 paper: PredRNN: Recurrent Neural Networks for Predictive Learning Using Spatiotemporal LSTMs.

nihaomiao /CVPR23_LFDM

The pytorch implementation of our CVPR 2023 paper "Conditional Image-to-Video Generation with Latent Flow Diffusion Models"

alexlee-gk /video_prediction

Stochastic Adversarial Video Prediction

thuml /RLVR-World

《RLVR-World：带强化学习的世界模型训练》（NeurIPS 2025）官方仓库，https://arxiv.org/abs/2505.13934

Westlake-AI /MogaNet

[ICLR 2024] MogaNet：高效多阶门控聚合网络

Yunbo426 /predrnn-pp

Code release for "PredRNN++: Towards A Resolution of the Deep-in-Time Dilemma in Spatiotemporal Predictive Learning" (ICML 2018)

google /e3d_lstm

e3d-lstm; Eidetic 3D LSTM A Model for Video Prediction and Beyond

thuml /iVideoGPT

Official repository for "iVideoGPT: Interactive VideoGPTs are Scalable World Models" (NeurIPS 2024), https://arxiv.org/abs/2405.15223

Yunbo426 /MIM

Code release for "Memory In Memory: A Predictive Neural Network for Learning Higher-Order Non-Stationarity from Spatiotemporal Dynamics" (CVPR 2019)

PRBonn /point-cloud-prediction

Self-supervised Point Cloud Prediction Using 3D Spatio-temporal Convolutional Networks

taldatech /lpwm

[ICLR 2026 Oral] Latent Particle World Models official repository

OpenRaiser /awesome-world-model-evolution

精选的研究论文、模型和资源合集，追溯从专业模型到统一世界模型的演变。

pairlab /SlotFormer

Code release for ICLR 2023 paper: SlotFormer on object-centric dynamics models

AlexanderMelde /SPHAR-Dataset

Surveillance Perspective Human Action Recognition Dataset: 7759 Videos from 14 Action Classes, aggregated from multiple sources, all cropped spatio-temporally and filmed from a surveillance-camera like position.

holmdk /Video-Prediction-using-PyTorch

Video Predicting using ConvLSTM and pytorch