ARGOMENTI PER PROGETTI E TESI
Last update 20/jan/2021
- [ML, THESIS] Action recognition: automatic identification of actions and goals of one or more humans from a series of observations on their actions and the environmental conditions (Deep Learning)
- [ML, THESIS] Sign language identification: evaluation of recent advances in hand-body pose estimator for the identification of the alphabet of sign language (DL).
- [ML] Transfer learning: evaluation transfer learning technologies by collecting photos from online image sharing services (e.g., Pinterest) and training DL-based approaches on a GPU-based workstation.
- [ML,THESIS] Behavior analysis in urban areas using Machine Learning and Computer Vision techniques: tirocinio presso azienda. (DL, activity recognition, pose estimation)
- [THESIS] Heart rate and SpO2 estimation based on Infrared Camera during the night.: tirocinio presso azienda. (ML,DL, time series analsysi)
- [THESIS] Face detection for infant safety: tirocinio presso azienda. (DL, activity recognition, pose estimation)
- [THESIS] Road luminance: tirocinio presso azienda. (ML,D)
- [THESIS] Body Temperature estimation: tirocinio presso azienda. (ML,DL)
Health recommender Systems
- [ML, THESIS] Activity & fitness trackers: analysis of a large dataset of daily-sampled signals obtained from fitness trackers for (1) Time series analysis for behavior identification (2) Correlation analysis (3) Health recommendation (Neural Networks, Deep Learning, Embeddings, Time series analysis)
MOOC (massive open online courses)
- [ML, SII, THESIS] Learning material identification: (1) collecting media from Youtube useful for learning purposes and course building. (2) Extraction of dialogues and text from media content. (3) Extraction of topics from media and (4) defining relationships between content, such as A is prerequisite of B, or A explains concepts cited by B (Natural language processing, Machine learning, Deep learning, embeddings, Audio analysis, computer vision).
Natural Language Processing
- [ML, SII] Automatic text generation using Deep Learning: evaluating recent DL-based technologies for the automatic generation of human-readable text (Natural language processing, DL BERT, Open AI GPT-2 and GPT-3)
- [ML, SII] Sentiment analysis/Opinion mining: experimenting recent DL-based technologies for sentiment and opinion mining (Natural language processing, DL BERT)
- [ML, THESIS] Prediction: evaluation of recent advances in prediction from time-series analysis in specific contexts, e.g., stocks market, traffic, etc (ML, Time series analysis)
- [ML, THESIS] Stock prediction: identification of relevant correlations between various sources and how stocks are moving in real-time (ML, Time series analysis)
Social networks analysis
- [ML, SII, THESIS] Intelligent crawling on Twitter via API: development of IA-based approaches for crawling the social networks and filtering relevant and personalized content for the user. Agent-based approach that makes use of Twitter APIs to collect data (ML, DL, Natural language processing).
- [ML, SII, THESIS] Community identification: recent advances in graph analyses for the identification of user communities on large social networks by using a GPU-based workstation (social network analysis, embeddings).
- [ML, SII, THESIS] Detection of tourist attraction from Instagram and similar services (DL, ML)
- [ML, SII, THESIS] User profiling: analysis of user behavior for inferring specific dimensions of the characteristics of the user, e.g., personality, emotions evoked by movies, sarcasm, aggressive intent, rumors, humor or offensive content, etc.
Online ML-based competitions
Per altre informazioni su tirocini e tesi magistrali realizzate presso Aziende, contattare il docente.