Figure 9
Web interface of makesense.ai (Skalski, 2019 ) with a subset of 400 images within the large X-ray scattering data set collected on beamline 7.3.3 at the ALS (Hexemer et al., 2010 ). The left-hand sidebar displays a set of thumbnails from the data set, while the center of the interface provides a full-screen view of the selected thumbnail image. The right-hand sidebar includes the label options available for assignment to the selected image. Its navigation bar offers a set of actions from which we can highlight (i) Run AI Locally, allowing users to upload a locally trained model for object detection purposes, with support for models such as YOLOv5 (Jocher, 2020 ), COCO SSD (Liu et al., 2015 ; Lin et al., 2014 ) and PoseNet (Kendall et al., 2015 ); and (ii) Connect AI Server, enabling users to run a remote model by connecting to a specified Roboflow (Dwyer et al., 2024 ) inference endpoint. |