Corporate Social Responsibility (CSR) reports serve as a formal medium through which companies communicate their environmental, social, and governance (ESG) initiatives, demonstrating long-term value creation for the environment, society, and the business itself. Among the various elements in these reports, images particularly those on the cover pages play a critical role in visually conveying the sustainability concerns that companies prioritize, thereby influencing stakeholder perceptions.
This thesis explores the use of image categories in the cover pages of sustainability reports to identify the messages or themes being communicated, in relation to sector-specific sustainability concerns defined by the Sustainability Accounting Standards Board (SASB). Furthermore, it investigates the alignment between the visual sustainability messages portrayed through cover images and the actual textual content of the reports. The analysis leverages the YOLOv5m object detection model for image classification and applies Natural Language Processing (NLP) techniques using Python to analyze the textual data.