photonix
photonix copied to clipboard
Use large "thumbnailed" images for classification instead of originals
I recently realised that face and object analysis is performed on the full, original image. Whilst this is good in some ways (able to recognise tiny faces) it means that a lot more memory is used and some images types (such as raw) may not be analysed. There may be other classifiers that are using the full image so check all.