IMPLEMENTATION OF BACKEND APPLICATION AS INTEGRATOR OF MACHINE LEARNING MODEL AND REAL TIME OBJECT DETECTION MOBILE APPLICATION FOR LOW VISION
Low vision is a condition in where a person's visual function decreases and this condition cannot be assisted with optical devices, making it difficult for the sufferer to carry out daily activities. The development of an object detection mobile application that helps people with low vision rec...
Saved in:
| Main Author: | |
|---|---|
| Format: | Final Project |
| Language: | Indonesia |
| Online Access: | https://digilib.itb.ac.id/gdl/view/69108 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Institution: | Institut Teknologi Bandung |
| Language: | Indonesia |
| Summary: | Low vision is a condition in where a person's visual function decreases and this condition cannot be assisted with optical devices, making it difficult for the sufferer to carry out daily activities. The development of an object detection mobile application that helps people with low vision recognize objects around them has detection limitations, which are only for objects that have been defined at the beginning. The development of the backend integrator application is intended to be able to connect the object detection model and mobile application so that the model's accuracy can continue to increase periodically. The backend integrator application consists of two components, namely an API backend application that provides a REST-based API to be able to provide object detection models to mobile applications and receive user-detected object images and a backend scheduler application to schedule periodic model retraining.
In designing the backend integrator application, the right architecture and technology are determined and the schema and entity features are determined in the database. The implementation of the backend API application includes the development of the APIs needed for CRUD operation of object image data and object detection models. The implementation of the backend scheduler application includes the construction of a scheduler with CRON configuration for retraining each model.
The result of this final project is the connection between the mobile application and object detection model efficiently. Periodic model training can generate new models that can be placed in the mobile application to update the previous model. Increasing the accuracy of model training has an impact on improving user experience for users. |
|---|