Fast and Automatic Object Registration for Human-Robot Collaboration in Industrial Manufacturing

TL;DR

This paper introduces a rapid, automated object detection retraining framework for human-robot collaboration in industrial settings, utilizing FPGA edge inference and a novel loss to improve open world recognition.

Contribution

It presents an end-to-end system for quick model updates and a new loss function to reduce false positives in recognizing unknown objects.

Findings

Significantly reduces false positives for unknown objects

Enables on-site model retraining with minimal human intervention

Achieves fast inference on FPGA edge devices

Abstract

We present an end-to-end framework for fast retraining of object detection models in human-robot-collaboration. Our Faster R-CNN based setup covers the whole workflow of automatic image generation and labeling, model retraining on-site as well as inference on a FPGA edge device. The intervention of a human operator reduces to providing the new object together with its label and starting the training process. Moreover, we present a new loss, the intraspread-objectosphere loss, to tackle the problem of open world recognition. Though it fails to completely solve the problem, it significantly reduces the number of false positive detections of unknown objects.