Deep Learning Based Racing Bib Number Detection And Recognition

Choi, Li Jia (2019) Deep Learning Based Racing Bib Number Detection And Recognition. Project Report. UTeM, Melaka. (Submitted)

Text (24 Pages) Deep learning based racing bib number detection and recognition.pdf - Submitted Version Download (714kB)

Abstract

Nowadays, healthy lifestyle trends are more prominent globally around the world. There are numerous number of marathon running race events that have been held and inspired interest among peoples of different ages, genders and countries. Such diversified truths increase more difficulties to comprehending large number of marathon images since such process is often done manually. Therefore, a deep learning based racing bib number (RBN) localization and recognition for marathon running races is implemented. RBN is the identification tag number wore by each runner during marathon running races. The architecture pipeline is consists of two phases. Two different networks are utilized which are You Only Look Once version 3 (YOLOv3) and Convolutional Recurrent Neural Network (CRNN). During first phase, YOLOv3 consists of single convolutional network that used to predict RBN by multiple bounding boxes and class probabilities of boxes. Next the RBN detected can be parsed into CRNN to undergo RBN recognition. For second phase, CRNN is used to output a label sequence for each input image and then selecting the label sequence that has the highest probabilities. As a result, CRNN will output the contents of RBN detected. All of the experimental results including of mean average precision (mAP) and edit distance have been analysed and evaluated in the project thesis.

Actions (login required)

View Item