These are chat archives for beniz/deepdetect

21st
Mar 2018
dgtlmoon
@dgtlmoon
Mar 21 2018 08:06
@thebluehue__twitter i had no problems running the examples on paperspace.com , I've even written a recipe here gist.github.com/dgtlmoon/f8fdb82ccaeb2b55f90a108129ecea23
@beniz I'm still working on an end-to-end example of what I'm aiming for before I email your team to ask about professional help (always better to have something demonstrable i think), however, my question about the index.ann... > so index.ann is 340Megabytes, for 750 images.. any strategies here? that's half a megabyte per image, and the image is 30kb :D
Emmanuel Benazera
@beniz
Mar 21 2018 08:20
Hi, index size per image depends on the size of the layer used for feature extraction. Default ggnet layer used in the example should be of size 1000 float I think.
dgtlmoon
@dgtlmoon
Mar 21 2018 09:31
@beniz ahuh interesting :) ok thanks
i wonder how small that size could be, and still be accurate enough to find similar images from a set of 500,000 :D
or 5,000 is fine too
Emmanuel Benazera
@beniz
Mar 21 2018 09:40
smaller size is usually better