Shops, supermarkets, and real estate dealers use marketing flyers in abundance to advertise the weekly and seasonal offers these days. The information available in such flyers is a good source marketing study. However, this information is not recorded in any central repository for future usage. This work focuses on the feasibility of using deep learning technology to detect objects from marketing flyers. The detection and recognition of objects from media files such as images is a prominent computer vision domain. Although previous investigations have used two-stage object detection techniques to solve the problem using models like Faster R-CNN, this work experiments with the usage of the state-of-the-art single-stage object detecting method YOLO in the object detection from marketing flyers. The work utilizes different training techniques of the YOLO algorithm and identifies the best one to use for detecting objects from marketing flyers.
Transfer learning and custom object training are the two methods of training YOLO. Transfer learning uses the pre-trained knowledge, while custom training is done with two different methodologies. One is by using a pre-annotated dataset like google open images. Another option is to collect representative data and manually annotate the object's position in it. Custom training with manual annotation achieved a mean average precision MAP of 98.56%. Hence it shows that single-stage object detection can be used to detect and classify objects from flyers provided to have representative datasets for training the model.