GaroBonneung

Overview

‘GaroBonneung’is a deep learning-based emotional routing application.

An emotional route is a path that considers the vibrancy, beauty, and stability of the street landscape, rather than just the shortest distance.

Collaborated with LX Corporation.

• Documentation:
• Technologies Used: | | |
• Dev. Environment: |
• Dev. Period: Mar. 2024 - Jun. 2024

My Contributions

1. Model Selection

• Selected a model trained with Tencent Map street landscape images, considering the license.

2. Score Prediction Pipeline

• Predicted scores of the street landscape images in all directions from a given location on the map.
• Normalized the predicted scores and tagged each node on the map.

3. Integrating Scores into the Routing Algorithm

   

• Incorporated the tagged scores into the routing algorithm.

4. Measuring Reliability through Surveys

   

• Conducted user satisfaction surveys to measure whether the routing results met the intended purpose.

Limitations

1. Unable to Select a Model Trained with Place-Pulse Dataset

• The Place-Pulse dataset, which uses GSV evaluated through crowdsourcing from various cities, could be more advantageous for predicting domestic street landscapes.
• Although we intended to select a model based on the Place-Pulse dataset, we adopted a model trained with Tencent Map street landscape images due to licensing issues.