AWS nova documentation change
Summary
Added documentation for new bounding box detection feature in Amazon Nova models
Security assessment
Documents new computer vision capability without any security implications
Diff
diff --git a/nova/latest/userguide/modalities-image.md b/nova/latest/userguide/modalities-image.md index 7447a1774..294bebcfa 100644 --- a//nova/latest/userguide/modalities-image.md +++ b//nova/latest/userguide/modalities-image.md @@ -5 +5 @@ -Image size informationImage to tokens conversion +Image size informationBounding box detectionImage to tokens conversion @@ -14,0 +15,4 @@ To provide the best possible results, Amazon Nova automatically rescales input i +## Bounding box detection + +The Amazon Nova Lite and Amazon Nova Pro models are trained to precisely detect bounding boxes within images. This capability can be valuable when the objective is to obtain the coordinates of a specific object of interest. The bounding box detection functionality of the Amazon Nova model makes it a suitable candidate for image grounding tasks, thereby enabling enhanced understanding of screen shots. The Amazon Nova model outputs bounding boxes on a scale of [0, 1000), and after these coordinates are obtained, they can be resized based on the image dimensions as a post-processing step. +