Deep Animator

Image animation given a driving video sequence and a source image.

Deep Animator is an implementation of First Order Motion Model for Image Animation by Aliaksandr Siarohin, Stéphane Lathuilière, Sergey Tulyakov, Elisa Ricci and Nicu Sebe.

The source code can be found here. This library is a simple transformation of the original code into a shell script for quick experimentation. This is also an educational effort for the writer.

Install

Run pip install deep-animator to install the library in your environment.

How to use

First you need to download the weights of the model here. Then just run the following command.

deep_animate <path_to_the_source_image> <path_to_the_driving_video> <path_to_yaml_conf> <path_to_model_weights>

• Example of source image here
• Example of driving video here

An example YAML file is given below:

model_params:
  common_params:
    num_kp: 10
    num_channels: 3
    estimate_jacobian: True
  kp_detector_params:
     temperature: 0.1
     block_expansion: 32
     max_features: 1024
     scale_factor: 0.25
     num_blocks: 5
  generator_params:
    block_expansion: 64
    max_features: 512
    num_down_blocks: 2
    num_bottleneck_blocks: 6
    estimate_occlusion_map: True
    dense_motion_params:
      block_expansion: 64
      max_features: 1024
      num_blocks: 5
      scale_factor: 0.25
  discriminator_params:
    scales: [1]
    block_expansion: 32
    max_features: 512
    num_blocks: 4
    sn: True