Video Prediction On Moving Mnist

Metrics

MAE

MSE

SSIM

Results

Performance results of various models on this benchmark

Model Name	MAE	MSE	SSIM	Paper Title
ConvLSTM	182.9	103.3	0.707	Convolutional LSTM Network: A Machine Learning Approach for Precipitation Nowcasting
MSPred	-	34.44	0.975	MSPred: Video Prediction at Multiple Spatio-Temporal Scales with Hierarchical Recurrent Networks
MAU	-	27.6	0.937	MAU: A Motion-Aware Unit for Video Prediction and Beyond
IAM4VPx5	49.2	15.3	0.966	Implicit Stacked Autoregressive Model for Video Prediction
PredFormer	41.96	11.62	0.9742	PredFormer: Transformers Are Effective Spatial-Temporal Predictive Learners
CrevNet+ST-LSTM	-	22.3	0.949	Efficient and Information-Preserving Future Frame Prediction and Beyond
VAN (SimVP 10x)	53.57	16.21	0.9646	MogaNet: Multi-order Gated Aggregation Network
Swin (SimVP 10x)	59.84	19.11	-	MogaNet: Multi-order Gated Aggregation Network
ConvMixer (SimVP 10x)	67.37	22.3	-	MogaNet: Multi-order Gated Aggregation Network
CrevNet+ConvLSTM	-	38.5	0.928	Efficient and Information-Preserving Future Frame Prediction and Beyond
Uniformer (SimVP 10x)	57.52	18.01	-	MogaNet: Multi-order Gated Aggregation Network
E3D-LSTM	86.4	41.3	0.910	Eidetic 3D LSTM: A Model for Video Prediction and Beyond
PredRNN-V2	-	48.4	0.891	PredRNN: A Recurrent Neural Network for Spatiotemporal Predictive Learning
MIM	116.5	52.0	0.874	Memory In Memory: A Predictive Neural Network for Learning Higher-Order Non-Stationarity from Spatiotemporal Dynamics
TAU	60.3	19.8	0.957	Temporal Attention Unit: Towards Efficient Spatiotemporal Predictive Learning
SwinLSTM	-	17.7	0.962	SwinLSTM: Improving Spatiotemporal Prediction Accuracy using Swin Transformer and LSTM
MLP-Mixer (SimVP 10x)	59.86	18.85	-	MogaNet: Multi-order Gated Aggregation Network
SA-ConvLSTM	94.7	43.9	0.913	Self-Attention ConvLSTM for Spatiotemporal Prediction
ViT (SimVP 10x)	61.65	19.74	0.9539	MogaNet: Multi-order Gated Aggregation Network
MogaNet (SimVP 10x)	51.84	15.67	0.9661	MogaNet: Multi-order Gated Aggregation Network

0 of 30 row(s) selected.