Malay traditional dessert image recognition using convolutional neural network / Yusra Syatirah Yusabri

This research was about Malay traditional dessert image recognition using Convolutional Neural Network (CNN). The extinction of many traditional desserts makes the current generations do not recognize the desserts. Some people might have seen the dessert but did not know the name. This is a problem because they could only describe the dessert. If they want to search for its names, it will become harder. To cater this problem, it is essential to provide a useful way for people to easily search about the desserts. Therefore, the objective of this project is to design and develop a Malay traditional dessert image recognition prototype using CNN technique. The methodology for this project involved five phases which were identifying problem, collecting data, designing and lastly developing, testing and fine tuning. For the data collection, there were 5 datasets with each dataset had 10 images for testing dataset and 20 images for training dataset. The project showed accuracy result for testing and training. The accuracy for training dataset is 0.5255 while 0.5978 for testing dataset. The result for testing the project is 57.14%. This conclude that the project is successful as it can run using the CNN technique. The accuracy result was not satisfying, but it can be improved in the future. As a conclusion, this project significantly helps general people to recognize and know more about the Malay traditional dessert.