Skip to main content
. 2023 Jun 27;13:10391. doi: 10.1038/s41598-023-34436-w

Figure 5.

Figure 5

Averaged statistics of the different models over the WF-6 dataset. This benchmark was run on a SpaceCloud device with an Intel X5 processor and a Intel Movidius Myriad2 chip. (a) Compression ratio of the final vectorised products (input size divided by output size) produced by the different ML models and baselines. (b–d) Intersection over Union, precision and recall of those models measured against the ground truth of the WorldFloods dataset12. (e) Total time (inference and vectorization) of the models running on different hardware. (f) Size of the weights in KB; this size is critical for updating the model after the satellite deployment.