kwatters

9 years 9 months ago

That's great, we should build an index repository of different objects so we can create training sets.

This could be a good use for Solr.  If we have a repository of images with metadata, we can index those, so we can run searches to return training sets both positive & negative examples....

 

Also, for reference , we can use the ImageNet dataset.  We should be able to borg something like this in.  I believe the data is freely available.

 

http://www.image-net.org/challenges/LSVRC/2014/

Banana Metadata!

http://www.image-net.org/challenges/LSVRC/2014/ui/api/bbox_api.php?type…

Wow .. I knew this was going to happen (Image Classifier Databases Springing Up) ..  I know Amazon has been collecting data for this purpose for a while..

That's very interesting !

I wonder if your Bananna Meta data is an OpenCV Haar classifier in JSON form (nice they moved format from xml !) :)