Just Say the Word: Annotation-Free Fine-Grained Object Counting
arxiv 2025
Moving beyond inference-time conditioning for fine-grained object counting by prompt tuning using synthetic data.
Adriano D'Alessandro
PhD Candidate (Computer Science)
I’m a computer vision researcher and PhD candidate living in Vancouver. My work focuses on object counting with limited data, and lately, I’ve been diving into using synthetic data from latent diffusion models to train counting models. I’m also into things like plant agriculture and biodiversity—basically, anywhere counting and nature intersect in interesting ways.
Publications
AFreeCA: Annotation-Free Counting for All
ECCV 2024
Utilizing latent diffusion models for annotation-free object counting across diverse categories.
Learning-to-Count by Learning-to-Rank
CVR 2023
Reducing the annotation burden for object counting by leveraging ranking supervision.
The human impact of data bias and the digital agricultural revolution
Handbook on the human impact of agriculture, Edward Elgar Publishing 2021
A analysis of big datasets and the ‘intelligent’ machines which use them in decision support systems for farmers.
Name2vec: Personal names embeddings
CCAI 2019
Creating name-embeddings for record linkage and document retrieval using Doc2Vec