Reading furniture and condition from listing photos
Supernova built the computer-vision pipeline behind ZipToZip's listing analysis. It collects hundreds of thousands of US property listings and reads each one from its photos: what furniture is in the rooms, whether the property is furnished or empty, and what condition it is in.
ZipToZip is a US real-estate analytics platform.
Investors and agents judge a property largely from its listing photos, but reviewing them by hand at that volume does not scale. ZipToZip needed to turn listing images into structured data automatically.
A pipeline that collects property listings from US sources on demand and pulls in their photos. For each listing it works through every photo, detects the furniture and objects in the rooms, classifies whether the property is furnished or empty, and reads its condition, then rolls the per-photo findings up into one structured record for the listing. Those records feed a dashboard where the team can filter by what the model found, review individual listings, and export the data into ZipToZip's own analytics. Collection and image analysis run as two separate services, so each scales on its own and an improved vision model can be re-run over past listings without touching the part that gathers them.