DEVELOPMENT OF A MOBILE APPLICATION TO ESTIMATE AGE RANGE BASED ON FACIAL PHOTOS AND FINE TUNING OF PRETRAINED CNN
Currently, computer vision technology is increasingly being used to help human life. One of them is the application of computer vision in facial recognition to estimate someone's age. The detection of a person's age which is implemented in this final project uses a pretrained CNN model...
Saved in:
| Main Author: | |
|---|---|
| Format: | Final Project |
| Language: | Indonesia |
| Online Access: | https://digilib.itb.ac.id/gdl/view/77864 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Institution: | Institut Teknologi Bandung |
| Language: | Indonesia |
| Summary: | Currently, computer vision technology is increasingly being used to help human
life. One of them is the application of computer vision in facial recognition to
estimate someone's age. The detection of a person's age which is implemented in
this final project uses a pretrained CNN model which has been fine tuned with
input in the form of facial photos. The system will group the face photos into a
certain age range. This grouping is intended to find out their age group. However,
detecting a person's age range accurately requires choosing the right dataset and
pretrained CNN model. Therefore, in this final project an experiment was carried
out on a pretrained CNN model to select a pretrained model that has a high
accuracy value in estimating age ranges. Furthermore, based on literature
studies that have been carried out, the current application of a person's age
detection system is still not practical. Some age detection systems that have been
developed still require a laptop or computer to run them. Apart from that, some
age detection system interfaces are still very simple, even using the command line
without a GUI. Thus, the development of a mobile application for age range
detection was then chosen in this final project by emphasizing an attractive user
interface and a pleasant user experience.
Based on the experimental results of the pretrained CNN model, MobileNet model
achieved the highest accuracy value by obtaining a value of 86.17% from 597 test
data. Thus, MobileNet model was chosen to be integrated into the mobile
application. Furthermore, based on the results of mobile application testing, the
system can fulfill all functional and non-functional requirements that have
previously been defined. However, the current system cannot be used offline. |
|---|