gifscom

gifscom / deepshotdetection / 0.2.0

README.md

Overview

This API provides highly accurate shot boundary detection (SBD). The goal of SBD is to split an edited video into consecutive frames which show a continuous progression of video, as shown in this illustration:

For obtaining high-performance results, our method uses a deep learning, more specifically a 3D convolutional neural network. Our approach has significantly better performance than existing SBD methods while being extremely fast. For more information on how it works, see our blog post.

Usage

The API takes a video url as input and returns a list of shots.

NOTE This API is currently in beta. If you have any issues or want to use private videos hosted on platforms such as Algorithmia or Google Cloud Storage or sources such as YouTube, please get in touch with us via* michael [at] gifs.com*.

Input

ParameterDescription
video-urlA public URL pointing to a video file

Output

ParameterDescription
shotsa list of shots of the form [start_time, end_time] in seconds

Gaps between consecutive shots occur for slow transitions such as fades or dissolves.

Example

Input:

"http://www.sample-videos.com/video/mp4/240/big_buck_bunny_240p_2mb.mp4"

Output:

[
  [0, 8.333333015441895],
  [8.399999618530273, 12.93333339691162],
  [13, 16],
  [16.066667556762695, 21.66666603088379],
  [21.733333587646484, 23.53333282470703],
  [23.600000381469727, 25.66666603088379],
  [25.733333587646484, 26.733333587646484]
]