fabrizzioalco / yolo-detection

Object detection in video and images using YOLOv3

yolo-detection

Python script using YOLOv3 algorithm and OpenCV to detect objects in images and video.

To get the helper files use the following commands:

wget https://pjreddie.com/media/files/yolov3.weights
wget https://github.com/pjreddie/darknet/blob/master/cfg/yolov3.cfg\?raw=true -O yolov3.cfg
wget https://github.com/pjreddie/darknet/blob/master/data/coco.names\?raw=true -O coco.names

To import the detectors:

from detectors import VideoDetector
from detectors import ImageDetector

The general process is:

from detectors import VideoDetector
d = VideoDetector(config_file='config.json')
# or d = VideoDetector(config_dict={})
"""
Optional kwargs:
- config_file: a JSON file with your configurations
- config_dict: a dictionary with your configurations
Just one is accepted, in the order mentioned.
"""
d.load_file('filename.mp4')
d.process()
d.write_json_to_file('filename.json') # optional: write the detection results to a JSON file.

Configurations accepted in the configuration file or dictionary:

{
  "model-cfg": "The path or filename of your yolov3.cfg",
  "model-weights": "The path or filename of your yolov3.weights",
  "model-classes": "The path of filename of your coco.names",
  "output-filename": "Suffix that is added to the output filename",
  "frame-skip": "[VIDEO] The number of frames to skip",
  "fps": "[VIDEO] FPS of the given video file",
  "confidence-threshold": "Minimum confidence value to keep the detected object",
  "nms-threshold": "Minimum non maximum supression value to keep the detected object in postprocessing",
  "input-width": "Width of network's input image",
  "input-height": "Height of network's input image"
}

Based from: https://github.com/spmallick/learnopencv/tree/master/ObjectDetection-YOLO