Object Recognition has recently become one of the most exciting fields in computer vision and AI. The ability of immediately recognizing all the objects in a scene seems to be no longer a secret of evolution. With the development of Convolutional Neural Network architectures, backed by big training data and advanced computing technology, a computer now can surpass human performance in object recognition task under some specific settings, such as face recognition.
I fell like whenever such an amazing thing happens; someone must tell the story of it. That is why this infographic was born. Its mission is to tell the modern history of object recognition in the most concise and engaging way. The story began since AlexNet won the ILSVRC 2012 competition and still being written. The infographic consists of two pages, where the first page summarizes important concepts while the second page sketches up the history. Every illustration has been remade to be more consistent and understandable. All the references have been cherry-picked so that you always know where to look for detail explanation. With that said, it will be hard to completely understand all of the concepts just by reading the infographic without prior knowledge about the field. If you find any trouble at understanding some concepts, leave me a comment, and I will be happy to be your explainer.
If the interest from the community is big enough, I can consider making a series of blog posts to go through and explain the whole infographic in detail.
How to print and read:
For best view on screen, you should download the pdf file. I attached the png file here just because medium does not support pdf format. The pdf file can be downloaded here.
The infographic is read best when it is printed on A3 paper. Download the pdf file and choose auto-center and scale 130% when printing to A3 paper. You can still print it on A4 paper, although the text will be pretty small. You should set the printing scale to 90% when printing to A4 paper.
The whole project is hosted here on Github.
Note: If you want to know more about the Receptive Field concept in CNNs, this post is for you.