Select Images for Training
When you train Media Server you must provide one training image for each image comparison reference.
A good training image for image comparison:
- should measure at least 400 pixels along the shortest dimension.
- contains only the part of the scene that you want to compare. If you take a reference image from a camera but only want to detect changes in part of the scene, you should crop the image before adding it to the image comparison database. When you run image comparison, you can set the
region
configuration parameter to limit analysis to the corresponding part of the ingested (source) media. - is not blurry.
- has not been converted from another file format to JPEG. Converting images to JPEG can introduce compression artifacts. If you have JPEG images, avoid making many rounds of edits. Each time that a new version of the image is saved as a JPEG, the image quality degrades.