SeeMore

A pytorch implementation of the paper "SeeMore: bidirectional spatio-temporal predictive model from the knowledge-transfer perspective". The code is based on PredRNN: A Recurrent Neural Network for Spatiotemporal Predictive Learning (TPAMI 2022).

Get Started

1. Install Python 3.6, PyTorch 1.9.0 for the main code.

2. Download data. This repo contains code for two datasets: the Moving Mnist dataset and the KTH action dataset.

3. Train and test the model. You can use the following bash script to train and test the model. The learned model will be saved in the --save_dir folder.

# Moving Mnist dataset
sh mnist_script/train_stage_1.sh
sh mnist_script/train_stage_2.sh
sh mnist_script/test.sh

# KTH action dataset
sh kth_script/train_stage_1.sh
sh kth_script/train_stage_2.sh
sh kth_script/test.sh