| Title |
Evaluating Store Image and Creating Positioning Maps Based on Deep Learning |
| Authors |
한유진(Han, Yoojin) ; 이현수(Lee, Hyunsoo) |
| DOI |
https://doi.org/10.5659/JAIK.2023.39.2.121 |
| Keywords |
Deep Learning; CNN; Image Classification; Interior Design; Store Image; Positioning Map |
| Abstract |
This study presents a deep learning approach to measuring a brand’s store image while generating positioning maps using social media data.
Store design and architecture were highlighted as effective communicators of brand identity and positioning, but the spatial environment had
been solely studied using traditional approaches such as surveys. This study adopted deep learning based CNN, which is an alternative
methodology for evaluating a brand’s store image and created a positioning map in terms of interior design. Two axes were set to create a
positioning map of style (X) and atmosphere (Y) that collected training data from Pinterest. Using the training dataset, this research employed
Inception-V3 to retrain this model to evaluate the interior design. Based on the retrained model, the interior images of coffee shop brands
were evaluated to determine each brand’s position and create a positioning map. Another positioning map was created based on a
conventional method via a survey to demonstrate the validity of this approach. The results demonstrated that a brand’s store image can be
trained and recognized using social data and deep learning technology. Additionally, brands’ relative positions and relationships can be
assessed through a deep learning framework; therefore, a brand positioning map can be created. Various applications of these approaches in
decision-making for brand store design, including the assessment of brand store positioning and redesigning stores were highlighted. Lastly,
this study suggests wider uses for social big data and deep learning technology in branding and architectural design. |