<p>Artificial Intelligence (AI) has been widely used in different domain such as self-driving, automated optical inspection and also detection of object locations for robotic pick and place operations. Although the current results of using AI in the mentioned fields are good, the biggest problem is that AI requires huge amount of data and labeling corresponding answers to allow sufficient training of AI model. Evidentially, these efforts still require lots of manpower. Once the quality of labeled answer is unstable, the trained AI model and the consequence results are inevitably altered. To resolve this issue, the auto annotation system is proposed in this paper with methods including (1) highly realistic model creates way with real texture, (2) domain randomization algorithm in the simulator to automatic generated huge and abundant images, and (3) light tracking algorithm to calculate the shielding effect of objects to offer each object different picking strategy label. It is found that with the mentioned methods, 10,000 various of images per hour can be generated in which each image can have multiple objects and each object can be labelled in different classes based on the different part of the object. Instance segmentation AI model can also be trained with the methods to verify the gaps between cyber and physical results, indicating that even when intersection over union (IOU) for each object reaches 70%, the mean average precision can reach 70%.</p>

A New Approach to Enhance Artificial Intelligence for Robot Picking System Using Auto Picking Point Annotation

Paper Proceedings: https://asmedigitalcollection.asme.org/ISPS/ISPS2021/volume/84799

ASME 2021 Information Storage and Processing Systems (ISPS2021)

Supporting the need among researchers worldwide for additional online and multimedia content, ASME now offers the ASME Conference Video Collection. This platform will collect and provide long-term access to conference presentation videos, encompassing all technology areas within ASME's technical conference and events program. About 30,000-40,000 engineers across all mechanical engineering and related disciplines participate in ASME conferences each year. This includes authors of approximately 6,000 to 8,000 technical papers published by ASME, the majority of whom also submit video presentations for our virtual conferences. The video collection will be available as an add-on to existing product packages or as a separate subscription.