fvmd 1.0.0

Frechet Video Motion Distance

FVMD

Fréchet video motion distance(FVMD) is a metric to evaluate the motion consistency of video generation.

🔨 Installation

Install with pip

pip install fvmd

🚀 Usage

Video Data Preparation

The input video sets can be either in .npz or .npy file formats with the shape [clips, frames, height, width, channel], or a folder with the following structure:

Folder/
|-- Clip1/
|   |-- Frame1.png/jpg
|   |-- Frame2.png/jpg
|   |-- ...
|
|-- Clip2/
|   |-- Frame1.png/jpg
|   |-- Frame2.png/jpg
|   |-- ...
|
|-- ...

Evaluate FVMD

To evaluate the FVMD between two video sets, you can run our script:

python -m fvmd --log_dir <log_directory> <path/to/gen_dataset> <path/to/gt_dataset>

You can alose use our FVMD in your Python code:

from fvmd import fvmd

fvmd_value = fvmd(log_dir=<log_directory>, 
                  gen_path=<path/to/gen_dataset>, 
                  gt_path=<path/to/gt_dataset>
                 )

Evaluate FVMD step by step

You can also run only some intermediate steps of FVMD.

Video Key Point Tracking

from fvmd import track_keypoints

velocity_gen, velocity_gt, acceleration_gen, acceleration_gt = keypoint_tracking(log_dir= < log_directory >,
gen_path = < path / to / gen_dataset >,
gt_path = < path / to / gt_dataset >
)

Extract motion feature from velocity/acceleration fields

from fvmd import calc_hist

motion_feature = calc_hist(vectors=<velocity_gen/velocity_gt/acceleration_gen/acceleration_gt>)

Compute FVMD from velocity/acceleration fields

from fvmd import calculate_fvmd_given_paths

results = calculate_fvmd_given_paths(gen_path=<directory/of/gen_velocity/acceleration_cache>, 
                                     gt_path=<directory/of/gt_velocity/acceleration_cache>
                                    )