Eidetic 3D LSTM: A Model for Video Prediction and Beyond

By Yunbo Wang, Lu Jiang, Ming-Hsuan Yang, Li-Jia Li, Mingsheng Long, Li Fei-Fei. International Conference on Learning Representations, 2019

439 citations recorded by the source corpus.

View Paper: https://www.semanticscholar.org/paper/82a646e1cb33124f672beba451f5039e9e32fb6d

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

Citations in the corpus

Earthformer: Exploring Space-Time Transformers for Earth System Forecasting (2022, methodology): “From top to bottom: input frames, target frames, predictions by Rainformer [3], UNet [43], ConvLSTM [36], PhyDNet [13], E3D-LSTM [44], PredRNN [45], Earthformer without using global vectors, Earthformer.”
SimVP: Simpler yet Better Video Prediction (2022, methodology): “Sixteen baselines are included, such as MCnet [62], ConvLSTM [71], SAVP, SAVP-VAE [30], VPN [28], DFN [25], fRNN [43], Znet [74], SV2P [3], PredRNN [67], VarNet [27], PredRNN++ [65], MSNET [31], E3d-LSTM [66] and STMFANet [26].”
Latent Video Transformer (2020, unknown): “Video prediction and unconditional video synthesis have been addressed for a long time and the solutions include recurrent models [15, 47, 48, 4], VAE-based models [8, 11, 25, 20, 9], autoregressive models [33, 39, 50, 22, 49, 19], normalizing ﬂows [24], GANs [27, 44, 35, 40, 1, 36, 7, 26] and…”
PredRNN: A Recurrent Neural Network for Spatiotemporal Predictive Learning (2021, methodology): “[9] enriched each RNN state by incorporating a time dimension that covers a short snippet, and proposed to model the dynamics using 3D convolutions and temporal self-attention.”
A review of convolutional neural networks in computer vision (2024, unknown): “Wang et al. (2019) presented a three-dimensional CNN architecture in conjunction with LSTM to distinguish various activities from video frames in order to address the issues of poor dynamic information obtained by the model, low prediction accuracy, and bad quality of the produced pictures.”
Disentangling Physical Dynamics From Unknown Factors for Unsupervised Video Prediction (2020, methodology): “Further works explore different architectural designs based on Recurrent Neural Networks (RNNs) [66, 64, 44, 67, 65] and 2D/3D ConvNets [40, 62, 50, 6].”
A Review on Deep Learning Techniques for Video Prediction (2020, methodology): “4: Representation of the 3D encoder-decoder architecture of E3d-LSTM [66].”
Joint Hand Motion and Interaction Hotspots Prediction from Egocentric Videos (2022, background): “On the other hand, predicting pixels for future frames [9, 44, 53, 84] is very challenging due to its high dimension outputs with large uncertainties.”