cscg / yolo-python-rtsp

Object detection using deep learning with Yolo, OpenCV and Python via Real Time Streaming Protocol (RTSP)

Object Detection with Yolo, OpenCV and Python via Real Time Streaming Protocol (RTSP)

Object detection using deep learning with Yolo, OpenCV and Python via Real Time Streaming Protocol (RTSP)

Recognized objects are stored in date seperated in folders per class for further training or face recognition.

OpenCV dnn module supports running inference on pre-trained deep learning models from popular frameworks like Caffe, Torch and TensorFlow.

When it comes to object detection, popular detection frameworks are

• YOLO
• SSD
• Faster R-CNN

Support for running YOLO/DarkNet has been added to OpenCV dnn module recently.

Dependencies

• opencv
• numpy
• imageio ( version 2.4.1 )

pip install numpy opencv-python imageio sudo pip3 install imageio==2.4.1

Note: Python 2.x is not supported

YOLO (You Only Look Once)

Download the pre-trained YOLO v3 weights file from this link or for tiny weights for slower machines link and place it in the current directory or you can directly download to the current directory in terminal using

$ wget https://pjreddie.com/media/files/yolov3.weights

$ wget https://pjreddie.com/media/files/yolov3-tiny.weights

Provided all the files are in the current directory, below command will apply object detection on the input image dog.jpg.

$ python yolo_opencv.py --input sampledata/commuters.mp4 --config cfg/yolov3.cfg --weights yolov3.weights --classes cfg/yolov3.txt

For RTSP simply put the RTSP URL as --input

$ python yolo_opencv.py --input rtsp://xxxxx:1935/live/sys.stream --framestart 100 --framelimit 100 --config cfg/yolov3-tiny.cfg --weights yolov3-tiny.weights --classes cfg/yolov3.txt

Arguments

parameter	type	description
input	String	/path/to/input/stream
outputfile	String	/path/to/outputfile
outputdir	String	/path/to/outputdir
framestart	Int	start detecting at frame x (int)
framelimit	Int	stop after x (int) frames and save the video in case of streams. 0 no limit
config	String	/path/to/config/file
weights	String	/path/to/weights/file
classes	String	/path/to/classes/file
invertcolor	Boolean	in case of BGR streams
fpsthrottle	Int	in case of slower machines to keep up with a stream

sample output :

Checkout the object detection implementation available in cvlib which enables detecting common objects in the context through a single function call detect_common_objects().

Credits

This project is based on Arun Ponnusamy's Object Detection OpenCV

Sample video footage from Videvo - Free Stock Video Footage