首页|Patent Application Titled 'Computer-Implemented Method For Generating (Training) Images' Published Online (USPTO 20240096069)
Patent Application Titled 'Computer-Implemented Method For Generating (Training) Images' Published Online (USPTO 20240096069)
扫码查看
点击上方二维码区域，可以放大扫码查看
原文链接
NETL
NSTL
Reporters obtained the following quote from the background information supplied by the inventors:“Machine learning models that perform object segmentation can be used within an automated roboticsystem to identify objects that an automated robotic manipulator can grasp, in order to fulfil an objecttranslation task su ch as removing an object from a container, or placing object inside a container. However,machine learning models that perform object segmentation are also sens itive to visual context. In practice,a large amount of training data is require d to train the machine learning models that implement objectsegmentation, or th at can be used at other stages of an automatic robotic manipulation pipeline. Acquisition of training data is time consuming and usually requires human interven tion.
NETL
NSTL
