Active Object Manipulation Facilitates Visual Object Learning: An Egocentric Vision Study

TL;DR

This study explores how active object manipulation and limited training data influence visual object learning in infants, highlighting the importance of hand control in supervision signals for effective recognition.

Contribution

It introduces a new perspective on infant visual learning by analyzing the role of hand manipulation and limited data in improving object recognition.

Findings

Supervision with hand manipulation outperforms without hands.

Learning is effective even with few training images.

Hand control enhances supervision signals during learning.

Abstract

Inspired by the remarkable ability of the infant visual learning system, a recent study collected first-person images from children to analyze the `training data' that they receive. We conduct a follow-up study that investigates two additional directions. First, given that infants can quickly learn to recognize a new object without much supervision (i.e. few-shot learning), we limit the number of training images. Second, we investigate how children control the supervision signals they receive during learning based on hand manipulation of objects. Our experimental results suggest that supervision with hand manipulation is better than without hands, and the trend is consistent even when a small number of images is available.