Objective
This project arose from a very specific need of the company MasOrange. There are cases in which suppliers send the cover and the segmented title separately. However, in most cases, only the title page is received. This poses a problem when adapting the content for the interface, such as the Orange TV website. The covers take up a lot of space and may not fit well with the style of the interface. On the other hand, having the title segmented, since it takes up less space, makes it easier to adapt it to the design and integrate it better, facilitating the personalisation of the content.
Among the existing text segmenters, none is trained or specialised in segmenting artistic text such as that which appears on this type of cover. The main difficulty is that the title can be presented in very different styles: thin letters, 3D outlines, low contrast with the background, overlapping elements or even slanted. This makes developing an automatic system that works well in all these cases really complicated.
The main objective of this project is to develop an automatic system capable of segmenting the title of these covers. The idea is that it should not depend on a specific style, but should be able to generalise. Furthermore, in case the automatic segmentation is not adequate, a manual alternative is offered to ensure that there is always a solution.
To solve this problem, a pipeline consisting of several modules has been developed. Text detection models have been used to locate the text in the image. Processing techniques are applied to improve the quality of these regions and generate prompts from them. These prompts are used as input to segmentation models that generate the title segmentation mask. Finally, LoRA is applied to adapt SAM to cover text segmentation.
On a set of 250 covers, the results obtained were very positive, with an IoU of 0.86. The system achieved segmentations very similar to those of ground truth, demonstrating good generalisability.
BACHELOR’S THESIS BY:
ANA ORTEGA MATEO
Academic Experience
- Computer Science and Engineering, Universidad Carlos III de Madrid (September 2021 – September 2025)
- Master in Cybersecurity, Universidad Europea (October 2025 – now)
Work Experience
- Machine Learning Researcher – Universidad Carlos III de Madrid in collaboration with Grupo MasOrange (September 2024 — June 2025)
- Cybersecurity Specialist, Mas Orange (September 2025 – now)
- Event hostess – Nargy (September 2023 — June 2025)
- Leisure and free time monitor (June summer 2023 — June summer 2024)
Technical skills
- Programming languages: Python, C/C++, SQL, HTML/CSS, JavaScript.
- Libraries: Pandas, OpenCV, Numpy, PyTorch, Keras, Sci-kit Learn.
- Cloud Platforms: Google Cloud.
- Frameworks: GitHub, GitLab.