Back
Similar todos
test pipeline choosing least blurry image out of every 120 frames for smaller datasets 
#mused
fail w/medium dataset from 360 video (~6,000 images) like small, but full requires too much memory, next strategy is render each input separately #mused
clean data & filter blurry images then start processing/aligning image datasets again #mused
Reduce image quality 
#curated
fix memory bug 
#interiorai by scaling input image smaller
try save money by reducing steps of upscaler from 18 to 9 and image gen from 40 to 30 without noticeable differences so then we save 1 minute of GPU time per image gen or $0.07 per photo or tens of thousands $ per month 
#photoai
load videos low res 
#skrrt
change image to higher res images
better image processing failure fallbacks in 
#readbot
maybe use a smaller screenshot frame 
#openform
resize slow loading images 
#codewithhugo
reduced image compression to improve quality of images on gocapture 
#meratek
test & fail 3d environment reconstruction from only 50 images instead of video #mused
