fix long time bug where photos WITHOUT faces in them couldn’t be finished because the pipeline broke at faceswap and face detect, now we just use the previous image from pipeline (preprocessed) and continue with that to upscaler or noiser #photoai

make concept for high res wide angle fix #photoai face detect face boundaries, square crop, send to GPU, stitch back face

clean data & filter blurry images then start processing/aligning image datasets again #mused

spends hours optimizing the brightness, contrast and blur of the auto segmented object mask so it works well for virtual staging and adding auto upscale #interiorai

Resumed applying artificial vision to the 150k+ photo collection #guregipuzkoa

test conversions of 360 images to rectilinear for making splats, about 80 v&h fov gives best result so far #mused

show preprocessed low res preview first, then do post processing in background, when done replace preproc with final pic for speed #photoai

see progress of moving all old images to R2 #avatarai now at 87% done

test new method eq 2 perspectives for processing 360 video frames #mused

render from acropolis small model & start processing data for full #mused

spent 4-5 hours post processing my video with CapCut - takes sooo much time! And I thought I had a setup which only needs minimal video editing

make new tour & process / align some data manually because perfectly symmetrical architecture 🤦🏻‍♂️ #spaceshare

new method for tighter control of transitions between 360 images for walking around #mused

training on full dataset is taking too long so export just the Kings Chamber 360s for testing #mused

Hit the halfway mark on Sicily editing - I explored a lot here so this particular batch of photos in 2022 is taking absolutely forever #admin

4h of photoshop #lab

Work on image snapshots for #hey

generate 360° interiors with #interiorai

spend 4 hours figuring out how to upscale better #photoai