A training-free layer that turns a flat object detector into a hierarchical,
open-set one. Each detection is classified by taxonomic abstraction: the most
specific level the evidence safely supports, or an explicit UNKNOWN OBSTACLE,
never a confident wrong leaf.
A flat detector returns one fixed class or nothing. On an untrained object (a
horse-drawn carriage, an overloaded truck) it must mislabel it or drop it, both
unsafe. HOWC abstracts up a taxonomy to a still-useful category
(… → Truck → Transport Vehicle → Vehicle), bounded by a per-branch safety
floor so it never collapses into a useless "Object"; below the floor it flags an
explicit UNKNOWN OBSTACLE with an inspectable decision path.
1pip install -r requirements.txt
2python app.py # Gradio UI: upload an image, see the taxonomy decision
3python app.py --share # same, but also prints a temporary public URL (~72h)
CC BY-NC 4.0, matching the paper.
