Posibility to train models within the summary detector for specific object recognition?

[dumitrescu5]

Hi, I was wondering if it's possible in ammico to improve detection of specific objects (e.g., certain flags) by training a model specifically to this purpose? Or are there other ways to achieve this? Thanks:-)

[iulusoy]

Possibly use the multimodal search (MultimodalSearch) instead of the SummaryDetector, can you provide an example of an object and some files that contain it?

[dumitrescu5]

Thanks! Yes, that seems the most promising way, but I got a bit stuck. I can check if there similarity between images depicting the object and other images in the data base, but for each example image I obtain a separate similarity score, and there doesn't seem to be the possibility to teach the tool that the example images are representations of the same object that should be searched for in the larger data base... Unless I overlooked something? So basically, I don't know how to make it go through this reasoning: this is bunch of images about an object --> this is the object I'm looking for --> search for it in the data base... Thanks again for any thoughts.