Introduction to objection detection using YOLO
Welcome to the world of object detection!
YOLO is like a hello world program for object detection using Deep learning. YOLO stands for "You only look once". It is a real time object detection model.
Check out Mark jay's awesome tutorials for good understanding and working of Yolo.
Mark jay GitHub.
Mark jay Youtube.
Yolo: Speed 45 FPS, Fairly accurate. Yolov2: Speed 155 FPS, Less accurate.
Yolo can be compared to FCNN rather than RCNN, It splites the images into MxM grids It is a fully functional convolutional neural network that passes an image once through the FCNN and determines the predections. A single convolutional network simultaneously predicts multiple bounding boxes and class probabilities for those boxes.
In this project we will explore
1. Use yolo pre-trained model and weights to test object detection on a existing Image.
2. Use yolo pre-trained model and weights to test object detection on a existing video.
3. Develop a custom model to detect a custom object by training a flavour of yolo model.
4. Test our model for the detection of custom trained object on existing images.
5. Test our model for the detection of custom trained object on existing videos.
6. Test our model for the detection of custom trained object on a realtime video using webcam.