3 ms·
Why stop at just office chairs? Why not detect all categorizable objects in every photo and rank based on that? Could also crop just the object detection regio
by MaheshNat 2y ago
Why stop at just office chairs? Why not detect all categorizable objects in every photo and rank based on that?
Could also crop just the object detection regions of each image, run those cropped images through CLIP/SigLIP, then UMAP and HDBSCAN to view a 2 or 3 dimensional latent space clustering of office chair types.. might reveal some info as to what kinds of chairs exist in what geographical regions. Could use a VLM to auto-tag each cluster given a couple images from each one. Could run PCA on the CLIP embeddings and have some sliders for each principal component.. maybe the first is chair color or size or whatever
much data = much fun
- nomad86 2y agoI tried to detect models of specific chairs, but it's very difficult. To train the AI model, you need many photos of the same type of chair, and on the internet, you can usually find only one stock photo of each chair.
- pinoy420 2y agoMake synthetic dataset with reposing
- myself248 2y agoThere's an Internet Movie Car Database, and independently, an Internet Movie Firearms Database. Surely there are more. I feel like they should be one database with object_type=car and object_type=firearm respectively. And then I can finally search by object_type=vacuum_cleaner and find out the wild-looking ball-shaped vacuum in that sci-fi movie whose name escapes me...
- NikkiA 2y agoThe classic orange 60s hoover ball? https://d3j17a2r8lnfte.cloudfront.net/mvh/2024/3/medium/bzSoUDvubVNwQbRS3hdznjcA.jpeg https://d3j17a2r8lnfte.cloudfront.net/mvh/2024/3/medium/bzSo... My grandparents had one in the 70s, it always amazed me.