Malay traditional musical instruments recognition using pre-trained convolutional neural network / Nur Khairunisa Mohd Ali

This research was about Malay Traditional Musical Instruments recognition. Currently, a website to retrieve musical instrument by using an image has been developed but only focuses on the western musical instrument. There is a website that contains information about Malay traditional musical instruments, but it only provides brief information about the instruments. In this paper, we propose a framework of deep learning for Malay Traditional Musical Instruments recognition by using Pre-Trained Convolutional Neural Network (CNN). The process that involved to build this prototype are collecting the dataset, importing libraries and splitting the dataset, applying a Pre-Trained CNN which is inception, and training the network. An intermodal dataset that contains ten classes of four categories of Musical Instruments used to train the network. A total of 4000 images were used in training and testing set respectively. All the images from each class were resizedto227 × 227 pixel. For retrieval, best results are achieved when class-based predictions are used. An average classification accuracy of 99.77% and a mean average precision of 0.69 is achieved for retrieval task. Finally, besides recognizing the uploaded image, this prototype also can uncover general knowledge and understanding of Malay traditional musical instruments