ANALISA PERBANDINGAN PERFORMA ALGORITMA CONVOLUTIONAL NEURAL NETWORK DALAM KLASIFIKASI GAMBAR ABJAD BAHASA ISYARAT INDONESIA
| dc.contributor.author | Kamandanu, Kamandanu | |
| dc.contributor.supervisor | Mahdiyah, Evfi | |
| dc.date.accessioned | 2022-11-15T08:43:53Z | |
| dc.date.available | 2022-11-15T08:43:53Z | |
| dc.date.issued | 2022-07 | |
| dc.description.abstract | Background clutter is a background in an image that has a lot of noise or objects, making it difficult to focus on one object. Generally, images with a cluttered background are more common than images with a green background. The purpose of this study is to see whether the background image affects the performance of the convolutional neural network algorithm in classifying the Indonesian Sign Language Alphabet (BISINDO). This deep learning model uses 2860 images of primary data for each background with a total dataset of 5720 images, the data that has been collected is divided into training, validation and testing, 3x3 filter size, and a learning rate of 0.001 and 50 epochs. of 0.983, validation of 0.823 and testing of 0.67 for model 1 (Green Background), while for model 2 (Background Clutter) the training accuracy is 0.971, validation is 0.529 and testing is 0.38. It can be concluded tha a pictures with background clutter affects the accuracy of the model. | en_US |
| dc.description.sponsorship | Fakultas Matematika dan Ilmu Pengetahuan Alam | en_US |
| dc.identifier.citation | Perpustakaan | en_US |
| dc.identifier.other | Elfitra | |
| dc.identifier.uri | https://repository.unri.ac.id/handle/123456789/10747 | |
| dc.language.iso | en | en_US |
| dc.publisher | Elfitra | en_US |
| dc.subject | BISINDO | en_US |
| dc.subject | CNN | en_US |
| dc.subject | Classification | en_US |
| dc.subject | Model | en_US |
| dc.subject | Clutter | en_US |
| dc.subject | Background | en_US |
| dc.title | ANALISA PERBANDINGAN PERFORMA ALGORITMA CONVOLUTIONAL NEURAL NETWORK DALAM KLASIFIKASI GAMBAR ABJAD BAHASA ISYARAT INDONESIA | en_US |
| dc.type | Article | en_US |