gif_maker.py

from PIL import Image
import numpy as np


'''
Paper: `Deep learning for precipitation nowcasting: a benchmark and a new model`
'''
def save_gif(single_seq, fname):
    """Save a single gif consisting of image sequence in single_seq to fname."""
    img_seq = [Image.fromarray(img.astype(np.float32) * 255, 'F').convert("L") for img in single_seq]
    img = img_seq[0]
    img.save(fname, save_all=True, append_images=img_seq[1:])

def save_gifs(seq, prefix):
    """Save several gifs.

    Args:
      seq: Shape (num_gifs, IMG_SIZE, IMG_SIZE)
      prefix: prefix-idx.gif will be the final filename.
    """

    for idx, single_seq in enumerate(seq):
        save_gif(single_seq, "{}-{}.gif".format(prefix, idx))


if __name__ == '__main__':
    
    data = np.random.randn(1, 10, 96, 96) # (batch_size, seq_len, H, W)
    save_gifs(data, 'zyl')