Send me an email

Hi I'm Alberto

I'm a graduate Computer Engineering student, with a strong passion for Artificial Intelligence. Based in Rovigo, Italy but my heart will be forever in my Erasmus home Prague, Czech Republic.
Available to move anywhere in the world.

• Machine Learning:

• Robotics:

• App Programming:

• OOP:

• Web Development:

An Empirical Study on Ensemble of Segmentation Approaches: (🔗Code | 📄Published Paper)

• Empirical search of the best ensemble architecture to perform colorectal polyps segmentation
• Tested ensembles with different backbones, data augmentation methods and loss functions
• Personally, I implemented several loss functions to create ensembles with greater diversity
• Achieved 0.893 Dice score
• Developed with MATLAB

Sentiment Analysis for Climate Change: (🔗Code | ✍️Blog Article)

• Sentiment Analysis over a tweets dataset regarding climate change with BERT and other Machine Learning techniques (e.g. SVM)
• Category 13 of the Sustainable Development Goals defined by the United Nations: Climate Action
• Achieved 0.86 F1-Score
• Developed with Tensorflow, Pandas, Scikit-learn, Python

Better Quality Embeddings for Node Classification Combining Classificiation with Link Prediction in Graph Neural Networks (🔗Code | 📄Paper)

• Implementation of Link Prediction as self-supervised pre-training technique in Graph Neural Networks for node classification.
• Tested with GCN, GAT and GraphSAGE (with mini-batch generation) with MLP for improved learning during Link Prediction task.
• Personally, I developed the whole training loop. My colleagues implemented grid search algorithms.
• Developed with Pytorch, Pytorch Geometric, Python

🌤️ Multi-dataset Weather Classification: (🔗Code | 📄Report)

• Image classification over weather images coming from multiple datasets.
• Usage of joint training for classes having similar features across different datasets.
• Handling of dataset and class imbalance.
• Usage of histograms with Linear Neural Networks for testing over fully unseen dataset.
• Achieved 0.994 Accuracy
• Developed with Pytorch, OpenCV, Python

🐢 Sea Turtle Face Detection for Ocean Conservation: (🔗Code | 📄Report)

• Object detection of turtles' heads with RetinaNet algorithm
• Category 14 of the Sustainable Development Goals defined by the United Nations: Life Below Water
• Achieved 0.878 IoU score
• Developed with Tensorflow, Pandas, Python

Grayscale to RGB for CNN training: (🔗Code | 📄Report)

• Conversion of 16 grayscale foraminifera images to a single image in RGB format
• The goal is to improve the classification accuracy of a CNN, the conversion is done with unsupervised learning techniques
• Achieved 0.805 accuracy
• Developed with MATLAB

Where Am I? (🔗Code | 📄Report in Italian)

• Android application to track in real-time the longitude, latitude and altitude of the device. Showing also the history of the past 5 minutes. Everything on a map as well.
• Developed with Kotlin

and many others... Just look at my repositories.