The following screenshot (figure copyright: Ros Girshick) shows the improvement in mean average precision over years for object detection on the images from the PASCAL VOC image dataset. As you can see, up to 2012, the performance for object detection started to stagnate and slow down a little bit. In 2013, the deep learning approaches came around and performance received a boost from that time onward, getting better and better over time:
Algorithms such as Region-based-CNN (for example, Faster/Mask R-CNN) and YOLO have been developed to improve the precision of object detection drastically using deep learning. In this recipe, we will discuss a couple of popular fully convolutional network models for object detection, one of them being YOLO (You Only Look Once). This provides a high accuracy rate compared to other algorithms and runs in real time...
