What do we learn from a large-scale study of pre-trained visual representations in sim and real environments?

TL;DR

This large-scale empirical study evaluates five pre-trained visual representations across multiple tasks, robots, and learning paradigms, revealing insights into their transferability from simulation to real-world applications.

Contribution

It provides the first comprehensive analysis of PVRs in real-world tasks, demonstrating their transferability and the impact of data augmentation and fine-tuning.

Findings

Simulation performance trends predict real-world outcomes.

Achieved zero-shot transfer in indoor ImageNav.

Data augmentation and fine-tuning improve real-world performance.

Abstract

We present a large empirical investigation on the use of pre-trained visual representations (PVRs) for training downstream policies that execute real-world tasks. Our study involves five different PVRs, each trained for five distinct manipulation or indoor navigation tasks. We performed this evaluation using three different robots and two different policy learning paradigms. From this effort, we can arrive at three insights: 1) the performance trends of PVRs in the simulation are generally indicative of their trends in the real world, 2) the use of PVRs enables a first-of-its-kind result with indoor ImageNav (zero-shot transfer to a held-out scene in the real world), and 3) the benefits from variations in PVRs, primarily data-augmentation and fine-tuning, also transfer to the real-world performance. See project website for additional details and visuals.