Classifying large images from SD


I’m working on a project that is designed to load 3MP jpegs from an SD card and classifies them. I already managed to do this on a arduino Portenta by only decoding the averages of each MCU of the JPEG and thus downscaling it with factor 8 on load. However i would love to move this application to micropython in order to have easily swapable models and use larger models that don’t fit the 2MB flash of the Portenta.

However my OpenMV H7 Plus crashes when try to convert the loaded image to rgb or grayscale and as i understand it is not possible to rescale it before conversion.

Would something like this be possible at all using OpenMV firmware?

This is just because of the jpeg decoder we have right now… there’s a PR pending to fix this online here: imlib: Enable hardware JPEG decoder. by kwagyeman · Pull Request #1604 · openmv/openmv · GitHub

Would it be possible to test this fix?