Encoder-Decoder Based Prediction
Comment:
It was not until when we were writing the final report that we realized that we mistakenly mixed the training data and test data together for the tennis set. So the frames generated below are actually not for test data, but for training data. Performance on real test data is presented in our final report. This is a big lesson for us from this project.
Raw Flow
Linear Interpolation
Ours
Ground Truth
Raw Flow
Linear Interpolation
Ours
Ground Truth
Sequence Prediction
2017 Stanford CS231N