Unofficially Tensorflow implementation of From Here to There: Video Inbetweening Using Direct 3D Convolutions (https://arxiv.org/pdf/1905.10240.pdf)
Still training...
Tensorflow
T = 16 and D = 128
input: (x_s, x_e), Gaussian noise vector u ∈ R^D
output:(x_s, xˆ1, . . . , xˆT −2, x_e)
./data/KTH/download.sh
python3 train_kth_multigpu.py --gpu 0 --batch_size 32 --lr 0.0001
Some codes are from here.