Papers
19
Total Citations
419
H-Index
13
About
Carlo Ciliberto is a researcher whose work sits at the intersection of machine learning, computer vision, and robotics, with a particular focus on enabling autonomous agents to perceive and interact intelligently with the world around them. Much of his most influential research has been conducted in the context of the iCub humanoid robot platform, where he has pioneered methods for robust object recognition, tactile exploration, and visual attention. His contributions to deep convolutional neural networks for robotic vision — including work on improving invariance for object identification and teaching robots to recognize objects with limited supervision — have garnered significant attention, collectively accumulating hundreds of citations. Ciliberto has also made notable strides in active perception, developing Gaussian process-based strategies for tactile surface reconstruction, and in motion detection, revisiting classical optical flow methods for real-time robotic applications. A recurring theme across his portfolio is the challenge of bridging the gap between controlled laboratory conditions and unconstrained real-world environments. His creation of the iCub World dataset, acquired through human-robot interaction, reflects a commitment to building practical, ecologically valid benchmarks. Ciliberto's body of work represents a sustained and impactful effort to bring reliable, learnable perception to physical robotic systems.
Research Focus
Key Achievements
Top Papers
- 1Active perception: Building objects' models using tactile exploration58 citations · 2016
- 2
- 3Are we done with object recognition? The iCub robot’s perspective46 citations · 2018
- 4Teaching iCub to recognize objects using deep convolutional neural networks35 citations · 2015
- 5iCub World: Friendly Robots Help Building Good Vision Data-Sets30 citations · 2013
- 6
- 7Weakly supervised strategies for natural object recognition in robotics23 citations · 2013
- 8
- 9
- 10