Nicola Fanelli

I am a PhD student in Computer Science & Mathematics in the Department of Computer Science at the University of Bari Aldo Moro, where I work on computer vision and deep learning, under the supervision of Prof. Giovanna Castellano and Prof. Gennaro Vessio.

I am currently pursuing a PhD funded by a PhD fellowship within the framework of the Italian "D.M. n. 118/23" under the PNRR, Mission 4, Component 1, Investment 4.1 on the PhD project "Analysis and Valorization of Digitized Artistic Heritage using Artificial Intelligence techniques". I am currently working in the CILab lab.

In 2023, I obtained my Master's degree in Computer Science (AI curriculum) from the University of Bari Aldo Moro, with a focus on machine learning. During the Master's program, I completed numerous ML projects related to computer vision and NLP, culminating in my thesis on automatic artwork captioning. I also collaborated for four months with the National Research Council of Italy, where I extended my BSc thesis work on text complexity assessment with machine learning.

Email  /  GitHub  /  Google Scholar  /  LinkedIn  /  StackOverflow  /  Medium  /  Twitter

News

Research

I'm interested in computer vision, multimodal deep learning (particularly vision and language), and generative models (MLLMs and diffusion models), especially in the context of artwork analysis.

I Dream My Painting: Connecting MLLMs and Diffusion Models via Prompt Generation for Text-Guided Multi-Mask Inpainting

We inpaint multiple image regions, each with a different text prompt, and generate multi-mask prompt suggestions.

Art2Mus: Bridging Visual Arts and Music through Cross-Modal Generation

We extend the AudioLDM2 architecture to generate music from artworks on a dataset of image-music pairings collected using ImageBind.

Label Anything: Multi-Class Few-Shot Semantic Segmentation with Visual Prompts

We present Label Anything, an innovative neural network architecture designed for few-shot semantic segmentation (FSS). Label Anything supports multi-class segmentation with points, boxes, or masks as prompts and relaxes multiple constraints in support set creation for FSS.

Converso: Improving LLM Chatbot Interfaces and Task Execution via Conversational Forms

We develop a fully-containerized architecture for creating LLM chatbots and improve their performances in data acquisition with conversational forms.

Exploring the Synergy Between Vision-Language Pretraining and ChatGPT for Artwork Captioning: A Preliminary Study

We explore caption generation for digitized artworks using a noisy dataset of LLM-generated descriptions. We introduce CLIPScore weighting to weigh the importance of each caption based on its quality to improve performances.

Design and source code from Jon Barron's website