Fast and Automatic Object Registration for Human-Robot Collaboration in Industrial Manufacturing
This work addresses the need for efficient object registration in industrial manufacturing settings, enabling quicker adaptation to new objects in human-robot collaboration, but it is incremental as it builds on existing Faster R-CNN methods with a new loss function.
The paper tackles the problem of fast retraining of object detection models for human-robot collaboration by presenting an end-to-end framework that automates image generation, labeling, and on-site model retraining, reducing human intervention to providing a new object and label. It introduces an intraspread-objectosphere loss to address open world recognition, significantly reducing false positive detections of unknown objects, though it does not fully solve the problem.
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.