From 52dbd947e710cc91d3309bc96de3802bd922154c Mon Sep 17 00:00:00 2001 From: Michael Levit Date: Thu, 17 Feb 2022 10:20:39 +0200 Subject: Updated readme file in object detection example. Signed-off-by: Michael Levit michaell@emza-vs.com Change-Id: If59b6bc63375c3f1172505684d75097ff526e32e (cherry picked from commit fba4e3a12b50d0ebb3804c6bf22cf688fcdaeee8) --- docs/use_cases/object_detection.md | 10 ++++++---- 1 file changed, 6 insertions(+), 4 deletions(-) diff --git a/docs/use_cases/object_detection.md b/docs/use_cases/object_detection.md index 8062325..1b74cf1 100644 --- a/docs/use_cases/object_detection.md +++ b/docs/use_cases/object_detection.md @@ -14,11 +14,13 @@ - [Running Object Detection](./object_detection.md#running-object-detection) ## Introduction - This document describes the process of setting up and running the Arm® *Ethos™-U* NPU Object Detection example. - -This use-case example solves the classical computer vision problem of Object Detection. The ML sample was developed -using the *YOLO Fastest* model that was trained on the *Wider* dataset. +Object Detection is a classical computer vision use case in which specific objects need to be identified and located +within a full frame. In this specific example the model was trained for face detection. The ML sample was developed +using the *YOLO Fastest* model. To adopt the model for low power / low memory systems the input images to the model +are monochrome images. The model was trained on the *Wider* dataset (after conversion from RGB to monochrome) +and on *Emza Visual-Sense* dataset [www.emza-vs.com](www.emza-vs.com). +The model makes detection faces in size of 20x20 pixels and above. Use-case code could be found in the following directory:[source/use_case/object_detection](../../source/use_case/object_detection). -- cgit v1.2.1