Francesco Di Felice
Papers
4
Total Citations
46
H-Index
3
About
Francesco Di Felice is a robotics researcher whose work lies at the intersection of computer vision and robotic manipulation, with a primary focus on advancing 6D object pose estimation. His most impactful contribution, the i2c-net framework (28 citations), introduces instance-level neural networks for monocular category-level 6D pose estimation, a critical capability for enabling robots to grasp objects reliably in cluttered, real-world environments under varying lighting conditions. Building on this, Di Felice has pioneered the use of Graph Neural Networks (GNNs) for one-shot imitation learning in pick-and-place tasks (11 citations), allowing robots to learn task-specific rules from synthetic demonstrations alone—a significant step toward more flexible and data-efficient robotic training. His most recent work, Zero123-6D (5+ citations), pushes the boundaries of generalizability by leveraging diffusion models for zero-shot novel view synthesis, enabling category-level pose estimation without task-specific fine-tuning. Collectively, Di Felice’s research addresses a fundamental bottleneck in robotics: the trade-off between precision and adaptability. By combining deep learning with geometric reasoning, he is helping to create robotic systems that can perceive and manipulate objects with minimal prior exposure, a key enabler for next-generation autonomous manufacturing and service robotics.
Research Focus
Key Achievements
Top Papers
- 1
- 2
- 3
- 4