SAPUTRA, VINCENTIUS KURNIAWAN INDRA (2025) COMPARISON BETWEEN EFFICIENTNETB0 AND RESNET50 ARCHITECTURE FOR PET FECES CLASSIFYING. S1 thesis, UNIVERSITAS KATOLIK SOEGIJAPRANATA SEMARANG.
|
Text
21.K1.0062-VINCENTIUS KURNIAWAN INDRA SAPUTRA-COVER_a.pdf Download (657kB) | Preview |
|
![[img]](http://repository.unika.ac.id/style/images/fileicons/text.png) |
Text
21.K1.0062-VINCENTIUS KURNIAWAN INDRA SAPUTRA-BAB_I_a.pdf Restricted to Registered users only Download (594kB) |
|
|
Text
21.K1.0062-VINCENTIUS KURNIAWAN INDRA SAPUTRA-BAB_II_a.pdf Restricted to Registered users only Download (637kB) |
|
|
Text
21.K1.0062-VINCENTIUS KURNIAWAN INDRA SAPUTRA-BAB_III_a.pdf Restricted to Registered users only Download (894kB) |
|
|
Text
21.K1.0062-VINCENTIUS KURNIAWAN INDRA SAPUTRA-BAB_IV_a.pdf Restricted to Registered users only Download (2MB) |
|
|
Text
21.K1.0062-VINCENTIUS KURNIAWAN INDRA SAPUTRA-BAB_V_a.pdf Restricted to Registered users only Download (593kB) |
|
|
Text
21.K1.0062-VINCENTIUS KURNIAWAN INDRA SAPUTRA-DAPUS_a.pdf Download (656kB) | Preview |
|
|
Text
21.K1.0062-VINCENTIUS KURNIAWAN INDRA SAPUTRA-LAMP_a.pdf Restricted to Registered users only Download (1MB) |
Abstract
The following is an analysis comparing the performance and accuracy between EfficientNetB0 and ResNet50 architectures used to classify pet feces into four categories of diare, lack-of-water, normal, and soft-poop. Auto-Orient and Grayscaling are preprocess techniques used in the research to boost performance while complementary with hyperparameter tuning and data augmentation. Using a dataset from Roboflow, the data set was split into training, validation and test set and the experiments were conducted. They produced the following results without Auto-Orient preprocessing for EfficientNetB0, where it reached 97.48% accuracy and without Grayscale for ResNet50, where it reached 96.14% accuracy. The results indicate that EfficientNetB0 is computationally efficient while ResNet50 is heavily dependent on color features for classification. Specifically, the study shows this means the current state of the art model, EfficientNetB0, is appropriate for resource-poor environments, while ResNet50 is appropriate for cases where color makes a significant impact.
| Item Type: | Thesis (S1) |
|---|---|
| Subjects: | 000 Computer Science, Information and General Works 000 Computer Science, Information and General Works > 004 Data processing & computer science |
| Divisions: | Faculty of Computer Science > Department of Informatics Engineering |
| Depositing User: | Mr Yosua Norman Rumondor |
| Date Deposited: | 03 Jul 2025 00:15 |
| Last Modified: | 03 Jul 2025 00:15 |
| URI: | http://repository.unika.ac.id/id/eprint/37319 |
| Keywords: | UNSPECIFIED |
Actions (login required)
 |
View Item |