
AI Blueprints
By :

In order to detect which photos on social media have logos and recognize which logos they are, we will develop a series of increasingly sophisticated DL neural networks. Ultimately, we will demonstrate two approaches: one using the Keras
library in the TensorFlow
platform, and one using YOLO in the Darknet
platform. We will write some Python code for the Keras
example, and we will use existing open source code for YOLO.
First, we create a straightforward deep network with several convolutional and pooling layers, followed by a fully connected (dense) network. We will use images from the FlickrLogos
dataset (Scalable Logo Recognition in Real-World Images, Stefan Romberg, Lluis Garcia Pueyo, Rainer Lienhart, Roelof van Zwol, ACM International Conference on Multimedia Retrieval 2011 (ICMR11), Trento, April 2011, http://www.multimedia-computing.de/flickrlogos/), specifically the version with 32 different kinds of logos. Later, with YOLO, we will use the version with 47 logos...