Spaces:
Running
on
CPU Upgrade
Apply for community grant: Academic project (storage)
We created OpenStreetView-5M, the first open-access dataset for visual geolocation at a global scale. Our dataset is composed of over 5 million images taken across 225 countries and territories. We also benchmark many sota computer vision models to see which design leads to good visual geolocation performance.
The paper introducing our dataset was accepted at CVPR 2024, see our
Project Page for more details.
We build here a small demo with 50 images, with the goals of: (i) introducing our dataset, (ii) showing the complexity of visual geolocation, (iii) get a better idea of the human performance. The predictions of the AIs are pre-recorded so we don't need any GPU. We would love however to have persistent storage, and if possible, more CPUs.
Don't hesitate if you have any questions!
Hi
@osv5m
, your Space seems to be fast on cpu-basic (maybe it's because it only use 50 images for now?) and each Space has 50GB of ephemeral storage (docs). So I'm wondering if your Space really needs the persistent storage and cpu-upgrade. Would be nice if you could provide more details on why your Space needs them and which persistent storage you need.
Hi,
Indeed, it is just 50 images for now, so we may not need more cpu. We are recording (anonymously) the performance of user whoplay the game and would like to use the scores to add an average human performance baseline for users to compare themself to. As far as I understand, this would require persistent storage (but not a lot, just a couple Gigs should suffice).
Thanks anyway for the terrific platform