Hi, i want to ask if we really need a tool for converting or transforming images to graphs, if so is there any libs or standards doing or building such a tool, thanks in advance
@amine.lakhder - not sure specifically what you mean? Images are at their core a bunch of pixels. Are you trying to do image recognition and (for example), say that the image is an adult human wearing a blue shirt holding a red flower?
Yes exactly, this is what i mean, so we can leverage this new representation of images as graphs to help us on other areas like scene graph generation (the example of yours) and other applications like segmentation..., So i'm asking if there any tools or libs that can help doing the transformation pipeline, thanks
Hi,
your requirement seems like the ones of NLP, but this time on image data. As far as I understood AI, there is no one fits all AI solution for Images. Image recognition in some cases seem to have intensive preprocessing steps before of the picture to make it consumable by AI...
For personal experience I like node-red. There you have some good out of the box nodes(standard) to build custom set of image recognition as well as neo4j nodes to create graph constructs based on your recognized image data.
Now all depends on your requirements about the infrastructure and architecture.
Other tools like Pentaho PDI, Apache HOP are ETL based tools capable of executing Image recognition tools (in the transform phase) and load the image.
btw. I find it sad, that apoc abendoned Stanford NLP path. I liked it, because Stanford NLP is open source AFAIK.
regards
Thomas
Thanks a lot for helping with this, hope i can find a solution soon.
Regards
Amine