Fig. 7

Cloud-based access to the Prop3D-20sf Dataset via HSDS. HSDS creates Service Nodes, which are containers that handle query requests from clients, and Data Nodes, which are containers that access the object storage in an efficient, distributed manner. The Prop3D-20sf dataset can be used as input to train an ML model either by accessing the data via a Python client library (h5pyd) or through our separate DeepUrfold Python package, which supplies PyTorch data loaders [12]. This illustration was adapted from one that can be found at the HSDS webpage (available under an Apache 2.0 license, which is compatible with CC-by\(-\)4.0)