This paper aims to clarify the relationship by way of example the title of the various concepts, to the learning process, came up with the bigger picture, it is clearly visible all the way from start to finish.
First, the background is set for general items are made of wood.
- python is a "basic carpentry Handbook";
- numpy can be said segments processed python, and is the "basic carpentry Handbook" is applied to a field, such as "carpentry - Arms articles";
- tensorflow / pytorch can be said segments processed numpy is - a section "Woodworking weapon chapter" in the "carpentry - weapon articles - toy guns."
- First need to have the barrel, clip, butt so the basic toy parts, then assembled these basic components has become a complete toy gun.
Of course, the toy gun is also divided into multiple, such as gun, pistol, rifle, laser gun, etc., so different toy gun (target detection, classification, prediction, etc.), the basic member (tensorflow, pytorch etc.) and a corresponding method of assembling guide (CNN, YOLO, etc.) are also different.
tensorflow, pytorch provide some good use of ready-made things, in order to achieve a particular purpose AI, many people put forward many different ways to take advantage of these ready-made things to achieve this purpose, each method has advantages and disadvantages ( the efficiency, accuracy, etc.), the methods (or also referred to theory, there is the original method, conventional method is a modification of the existing method) is CNN, YOLO like.
