Training an object detector is easy when you already have labelled boxes. If you're working on a new domain or task, you often don't.
I've been experimenting with using an agent to bootstrap the data instead:
– a zero-shot model labels a sample (@TIIuae Falcon-Perception, 0.6B,
Every illustration in this dataset now has an instance mask: 411,385 cut-outs from 115,293 Encyclopaedia Britannica pages, 1768–1929. Public domain clip art, still a work in progress.
Masks generated by a @roboflow RF-DETR seg model (29M params) fine-tuned on @huggingface Jobs
Uploaded a dataset of 115,293 illustrated pages from the Encyclopaedia Britannica, 1st edition (1768) to 14th (1929) to the Hub
huggingface.co/datasets/bigla…
Uploaded a dataset of 115,293 illustrated pages from the Encyclopaedia Britannica, 1st edition (1768) to 14th (1929) to the Hub
huggingface.co/datasets/bigla…
