| Title |
Enhancing Code Generation from Images via Attention Layers |
| DOI |
https://doi.org/10.5370/KIEEP.2023.72.3.179 |
| Keywords |
Attention Mechanism; Code Generation; Screen Images; Convolutional Neural Network (CNN); Domain-Specific Language (DSL) |
| Abstract |
The process of developing web pages involves a collaboration between web designers and website programmers. While non-IT professional designers create visually appealing screens, implementing the corresponding source code remains a laborious and time-consuming task for developers. To address this challenge and enhance developer productivity, this paper presents the Pix2code-ATT model, an extension of the existing Pix2code model, which incorporates an attention mechanism. This attention layer empowers the proposed model to focus on critical elements during the code generation process, thereby optimizing its overall performance and accuracy. To evaluate the effectiveness of Pix2code-ATT, extensive experiments have been conducted with synthetic datasets. The experimental results demonstrate the model's capability to automatically generate source code from screen images while achieving text code creation within an acceptable range of error ratios. |