In image classification an image with a single object is the focus and the task is to say what it contains. But when we look at the world around us, we carry out far more complex tasks. There are multiple overlapping objects, different backgrounds and we not only classify these different objects but also identify their boundaries, differences, and relations among them. This task fall under the name of object detection and instance segmentation.
Can CNNs help us with such complex tasks, identify different objects in the image and their boundaries?
Through this seminar we’ll cover the intuition behind some of the main techniques used in object detection and segmentation and see how they’ve evolved from one implementation to the next. From R-CNN to Mask-RCNN, YOLO and SSD-Mobilenet.
How many data do they need to be trained on? Are they capable to act in real time, in a small device and in the wild? We will also try to discover how they deal with variance in scale, traslation and rotation of objects through practical tests on some pre-trained models.