eyecu-bumblebee 0.4.0

Advanced pipelines for video datasets

Bumblebee

Bumblebee provides high level components to construct training pipelines for videos conveniently.

• Install

• Motivation

• Our Websites

• Examples

  • A Pipeline with basic elements
  • Using Manager API

• Team

• License

Install

pip install eyecu_bumblebee

Motivation

Everything should be made as simple as possible, but no simpler. - Albert Einstein

Our Websites

EyeCU Vision
EyeCU Future

Examples

A pipeline with basic elements

from bumblebee import *


if __name__ == "__main__":
    
    VIDEO_PATH = "/path/to/video.mp4"

    # Create a source
    file_stream = sources.FileStream(VIDEO_PATH)

    # Add an effect
    goto = effects.GoTo(file_stream)

    # Add some transformers
    data = transformers.GrayScale(file_stream)
    data = transformers.Normalization(data)

    END_OF_VIDEO = file_stream.get_duration()
    goto(END_OF_VIDEO)

    # Create a dataset
    single_frame = datasets.SingleFrame(data)

    last_frame = single_frame.read()

Using Manager API

from bumblebee import *


if __name__ == "__main__":
    
    # Create a training manager
    manager = managers.BinaryClassification(
        ["path/to/video_dir","path/to/another_dir"],
        ["path/to/labels"]
    )

    number_of_epochs = 300
    
    for epoch,(frame_no,frame,prob) in manager(number_of_epochs):
        # Use data stuff
        ...    

Team

This project is currently developed and maintained by ovuruska.

License

Bumblebee has MIT license. You can find further details in LICENSE.