TRISIAWAN, INGGIS KURNIA (2021) PENERAPAN MULTI-LABEL IMAGE CLASSIFICATION MENGGUNAKAN METODE CONVOLUTIONAL NEURAL NETWORK (CNN) UNTUK SORTIR BOTOL MINUMAN. S1 thesis, Universitas Mercu Buana Jakarta.
|
Text (HAL COVER)
1. COVER.pdf Download (30kB) | Preview |
|
|
Text (ABSTRAK)
5. ABSTRAK.pdf Download (50kB) | Preview |
|
|
Text (LEMBAR PERNYATAAN)
3. HALAMAN PERNYATAAN.pdf Download (600kB) | Preview |
|
|
Text (LEMBAR PENGESAHAN)
2. HALAMAN PENGESAHAN.pdf Download (189kB) | Preview |
|
|
Text (KATA PENGANTAR)
4. KATA PENGANTAR.pdf Download (91kB) | Preview |
|
|
Text (DAFTAR ISI)
6. DAFTAR ISI.pdf Download (55kB) | Preview |
|
|
Text (DAFTAR TABEL)
8. DAFTAR TABEL.pdf Download (43kB) | Preview |
|
|
Text (DAFTAR GAMBAR)
7. DAFTAR GAMBAR.pdf Download (66kB) | Preview |
|
![[img]](https://repository.mercubuana.ac.id/style/images/fileicons/text.png) |
Text (BAB I)
9. BAB I.pdf Restricted to Registered users only Download (66kB) |
|
|
Text (BAB II)
10. BAB II.pdf Restricted to Registered users only Download (353kB) |
|
|
Text (BAB III)
11. BAB III.pdf Restricted to Registered users only Download (841kB) |
|
|
Text (BAB IV)
12. BAB IV.pdf Restricted to Registered users only Download (368kB) |
|
|
Text (BAB V)
13. BAB V.pdf Restricted to Registered users only Download (90kB) |
|
|
Text (DAFTAR PUSTAKA)
14. DAFTAR PUSTAKA.pdf Restricted to Registered users only Download (106kB) |
|
|
Text (LAMPIRAN)
15. LAMPIRAN.pdf Restricted to Registered users only Download (44kB) |
Abstract
In Fast Moving Consumer Goods (FMCG) industry such as a water bottle, quality control plays an important role to maintain good products; therefore, a fast and reliable method is needed to analyze field data and provide information that can help to determine product quality. During production in the factory, a water bottle often has multiple defects due to errors in the machine, human interventions, etc. For example, miss or incorrectly installed caps and labels, as well as underfilled water on the bottle. All of those issues can reduce the quality of the product shipped. To overcome that, visual inspection becomes a mandatory technique in the production lines. Traditional water bottle inspection methods often require multiple cameras and devices to detect multiple defects on the product, each one is used for detecting a different type of issue. In this paper, a method called Multi-Label Image Classification with Convolutional Neural Network (CNN) as an algorithm is being used to detect multiple defects on the water bottle at once. There are six class labels i.e underfilled water, empty water, broken cap, missing cap, broken label, missing label, and each one represents a probability of the defects is present on the bottle. Several combination of feature learning layer and fully connected layer on the CNN are used to extract patterns and classifying the input picture. To facilitate related study, a large-scale water bottle picture dataset is collected, which is associated with six predefined labels. The experiments demonstrate that the proposed model can achieve 98,526% accuracy prediction when tested with new datasets, and 97,71% average accuracy during testing with 10-fold cross-validation. Keywords: Artificial Intelegance (AI), Multi-Label Image Classification, Convolutional Neural Network (CNN), Water Bottle Inspection. Dalam industri Fast Moving Consumer Goods (FMCG) seperti air minum kemasan, quality control berperan penting untuk menjaga produk yang baik, oleh karena itu diperlukan metode yang cepat dan andal untuk menganalisis data lapangan dan memberikan informasi yang dapat membantu dalam menentukan kualitas produk. Selama proses produksi di pabrik, botol air minum sering kali memiliki beberapa cacat misalnya, tutup dan label yang tidak ada atau tidak terpasang dengan benar, serta air yang kurang terisi pada botol. Semua masalah tersebut dapat menurunkan kualitas produk yang dikirim ke konsumen. Untuk mengatasinya, inspeksi visual menjadi teknik wajib di lini produksi. Metode inspeksi visual tradisional sering kali memerlukan beberapa kamera dan perangkat untuk mendeteksi beberapa cacat pada produk, masing-masing digunakan untuk mendeteksi jenis masalah yang berbeda. Dalam penelitian ini, digunakan metode Multi-Label Image Classification dengan Convolutional Neural Network (CNN) sebagai algoritma untuk mendeteksi beberapa cacat pada botol air sekaligus. Terdapat enam class label yaitu isi kurang, isi kosong, tutup rusak, tutup tidak ada, label rusak, dan label tida ada. Masing-masing class label tersebut mewakili kemungkinan adanya cacat yang ada pada botol. Beberapa kombinasi feature learning layer dan fully connected layer digunakan untuk mengekstrak pola dan mengklasifikasikan gambar masukan. Untuk memfasilitasi penelitian ini, dikumpulan dataset gambar botol air dengan skala besar, dimana didalam datset tersebut merepresentasikan enam class label yang telah ditentukan. Saat diuji dengan dataset baru, model CNN mendapatkan hasil akurasi prediksi 98,526%, dan mendapat rata-rata akurasi sebesar 97,71% ketika diuji dengan 10-fold cross validation. Kata kunci: Artificial Intelegance (AI), Multi-Label Image Classification, Convolutional Neural Network (CNN), Sortir botol minuman.
Actions (login required)
 |
View Item |
