FE - DI | Dissertações de Mestrado e Teses de Doutoramento
Permanent URI for this collection
Browse
Browsing FE - DI | Dissertações de Mestrado e Teses de Doutoramento by Author "Abreu, Raquel Filipa Lebre de"
Now showing 1 - 1 of 1
Results Per Page
Sort Options
- Contrastive Learning for Visual Object RecognitionPublication . Abreu, Raquel Filipa Lebre de; Lopes, Vasco Ferrinho; Neves, João Carlos RaposoContrastive Learning has emerged as a key technique in machine learning, especially in the context of self-supervised learning. By using contrastive objectives, these models learn to approximate representations of similar data points in latent space, while distancing those that are distinct, promoting efficient discrimination between patterns. This approach has demonstrated a significant impact in several areas, including computer vision and natural language processing. This dissertation investigates the topic of Contrastive Learning, with a focus on visual object recognition. Specifically, the main contrastive methods are explored, with the aim of identifying limitations that can be addressed and improved. Various solutions have been proposed with the aim of overcoming the limitations found, which will be explored throughout this document. The results obtained varied in terms of performance: some of the solutions showed lower performance compared to the original results, while others came very close to the initial results, but did not surpass them. However, one of the solutions outperformed the base methods, demonstrating the possibility of further improvements through continuous adjustments and improvements.