Models and Data
Meadow Dataset
An open, licence clean dataset of twelve thousand annotated plant images for agritech models.
A compact vision model fine tuned to read handwritten shipping labels in poor light.
Loom Vision is deliberately small so it can run on the scanning device instead of a server. The gain came from data rather than architecture: eighty thousand labels photographed in real warehouses, in real light, with real smudges. Accuracy on the messiest tenth of the set improved by thirty one points.
An open, licence clean dataset of twelve thousand annotated plant images for agritech models.