EZ-CV

v 0.1

Overview

EZ-CV is a library to enable wrapped prototyping of computer vision analytics for applications on images and videos. At present this library provides a typed schemas for input and output to obhect detectors that run on images or individual video frames. These schemas are used by the Streamer object (so named for it's eventual use in processing streaming video) to pass images to a registered object detection function and to process the results. The basic workflow is as follows:

1. Create an object detection function
2. Instantiate the Streamer object in the main function of the program. Pass in the object detection function as an argument
3. [Optional] Create and register an output function that acts on the output of the object detector (basic functions are included that display the images/frames with bounding box overlays or dump data to the command line.
4. [Optional] Create parameters that will be optional arguments to the main command. Typical uses would be model specific parameters or paths for files to load
5. [Optional] Create an init function which runs after arguments are parsed but before images/frames are processed.
6. Call the run() method on the streamer, passing it any parameters you created and an init function if required

EZ-CV will create a Click based CLI with 3 commands

1. image: which takes as argument an image file path and passes that to the object detector
2. video: which takes as argument a video file path and passes each frame of the video to the object detector
3. camera: which takes an optional argument for the camera ID and streams frames from the webcam to the object detector

Installation

$ git clone https://github.com/PVjammer/ezcv.git
$ cd ezcv
$ pip install .

TODO

• Add command to stand up service
• Add support for other image/video functions