Detection

This class provide the function for obtaining the detected images.

Link to the class

Attributes

Attributes	Type	Description
categories	dict	Count all the detected objects.
img	numpy.array	Origin image get from the camera.
size	tuple	Size of image.
frame	numpy.array	Image with bounding boxes and categories.

Constructor

__init__(self, google_kit_json_path: str, categories: list, size: tuple = (640, 480), max_results: int = 10, camera=0)

Create an instance from Detection class.

• Args
  • google_kit_json_path: Path for the Google API JSON file. To get the API file, please refer to this page .
  • categories: Categories to be detected. All supported types can be found in here.
  • size: Size of the images and frames.
  • max_results: Maximum results for the response. It won't absolutely contain the wanted objects listed in categories. The categories being selected is controlled by the Google Server.
  • camera: Camera to be used. The backend function was using cv2.VideoCapture in OpenCV. Please refer to this link for all the valid input.

Method

start(self)

Start the object detection and open the camera.

end(self)

End the detection and release the camera.

Logic

Below is a flow chart that shows the logic behind the class.

Example

import gcdetection
import cv2

# Create detected class and start the detection
detection_class = gcdetection.Detection(google_kit_json_path="/path/to/json/file",
                                        categories=["Person", "Book"],
                                        size=(640, 480),
                                        max_results=10,
                                        camera=0)
detection_class.start()

# Show the original images and detected images
while True:
    # Obtain images
    ori_img = detection_class.img
    detected_img = detection_class.frame

    # Wait for key
    key = cv2.waitKey(30) & 0xff
    if key == 27: # ESC
        break

    # Show the images
    if ori_img is not None:
        cv2.imshow("Original image", ori_img)
    if detected_img is not None:
        cv2.imshow("Detected image", detected_img)

# Release the resources
detection_class.end()
cv2.destroyAllWindows()

For more example, please visit the example repository.