Graph-based Vision-Language Model for Multi-Modal Retrieval of Fashion Products

Artificial Intelligence is reshaping the field of fashion industry in different ways. E-commerce retailers exploit their data through AI to enhance their search engines, make outfit suggestions and forecast the success of a specific fashion product. However, it is a challenging endeavour as the data they possess is huge, complex and multi-modal. The most common way to search for fashion products online is by matching keywords with phrases in the product's description which are often cluttered, inadequate and differ across collections and sellers. A customer may also browse an online store's taxonomy, although this is time-consuming and doesn't guarantee relevant items. With the advent of Deep Learning architectures, particularly Vision-Language models, ad-hoc solutions have been proposed to model both the product image and description to solve this problems. However, the suggested solutions do not exploit effectively the semantic or syntactic information of these modalities, and the unique qualities and relations of clothing items. In this work of thesis, a novel approach is proposed to address this issues, which aims to model and process images and text descriptions as graphs in order to exploit the relations inside and between each modality and employs specific techniques to extract syntactic and semantic information. The results obtained show promising performances on different tasks when compared to the present state-of-the-art deep learning architectures.

Characterization Of Single Use Reactor Used For The Manufacturing Of Formulated Products Within The Pharmaceutical Industry

Mixing is a fundamental unit operation in the pharmaceutical industry to ensure consistent product quality across different batches. It is usually carried out in mechanically stirred tanks, with a large variety of designs according to the process requirements. A key aspect of pharmaceutical manufacturing is the extensive and meticulous cleaning of the vessels between runs to prevent the risk of contamination. Single-use reactors represent an increasing trend in the industry since they do not require cleaning and sterilization, reducing the need for utilities such as steam to sterilize equipment and the time between production batches. In contrast to traditional stainless steel vessels, single-use reactors consist of a plastic bag used as a vessel and disposed of after use. This thesis aims to characterize the fluid dynamics features and the mixing performance of a commercially available single-use reactor. The characterization employs a combination of various experimental techniques. The analysis starts with the visual observation of the liquid behavior inside the vessel, focusing on the vortex shape evolution at different impeller speeds. The power consumption is then measured using a torque meter to quantify the power number. Particle Image Velocimetry (PIV) is employed to investigate local fluid dynamics properties such as mean flow field and mean and rms velocity profiles. The same experimental setup of PIV is exploited for another optical measurement technique, the Planar Laser-Induced Fluorescence (PLIF). The PLIF measurements complete the characterization of the reactor with the qualitative visualization of the turbulent flow and the quantitative assessment of the system performance through the mixing time. The results confirm good mixing performances for the single-use reactor over the investigated impeller speeds and reveal that the filling volume plays a significant role in the fluid dynamics of the system.