Tuning Modular Networks with Weighted Losses for Hand-Eye Coordination

TL;DR

This paper proposes an end-to-end fine-tuning approach using weighted losses to enhance hand-eye coordination in modular deep visuo-motor policies, significantly improving robotic reaching performance.

Contribution

It introduces a novel fine-tuning method with weighted losses specifically designed for modular networks trained independently.

Findings

Significant performance improvement in robotic planar reaching tasks

Effective end-to-end fine-tuning of modular visuo-motor policies

Enhanced coordination accuracy through weighted loss optimization

Abstract

This paper introduces an end-to-end fine-tuning method to improve hand-eye coordination in modular deep visuo-motor policies (modular networks) where each module is trained independently. Benefiting from weighted losses, the fine-tuning method significantly improves the performance of the policies for a robotic planar reaching task.