Skip to content

Latest commit

 

History

History
24 lines (20 loc) · 1.04 KB

README.md

File metadata and controls

24 lines (20 loc) · 1.04 KB

Custom Detector

Examples of custom detector for Local Motion. Custom detector must be Web API which takes image as input and returns normalized positions of bouding boxes.

Requirements of Custom Detector

Please refer to the sample source code of examples for details on how to implement it to meet the following requirement.

  • Custom Detector Web API has POST method
  • Local Motion sends an image of part of screenshot as base64 encoded string. Payload is supposed to have image_data field.
{
    "image_data": "base64_encoded_image_here"
}
  • Custom Detector Web API must return boxes which is list of bouding box position [xmin, ymin, xmax, ymax]. The each value must be normalized to 0~1. For example, if image width is 400px and xmin of a bouding box is 100px, xmin should be 0.25.
{
    "boxes": [[0, 0, 0.2, 0.3], [0.3, 0.4, 0.5, 0.45]]
}

List of Examples

  • Word Detector: Google Cloud Vision API is used to detect words for custom detector of Local Motion.
  • Whole Image Detector: Fixed single box position is returned.