Skip to content

T4B — Vision domain shift (multilayer FPN / U-Net)

Paper evidence: main.pdf · Block findings

Lemma: D4 · Stack: pytorch Nuisance key: domain_shift

Production change: Texture and scene style shift together; same label map.

Notebook (Run All, built-in demo): t04b-multilayer-vision.ipynb

pip install matching-pmh torch
# Open the notebook and Run All

What this task achieved (headline)

E1_multiscale rare-5 Cityscapes mIoU 30.75% (+11.1 pp vs B0 19.68%).

B0 E1 E1_multiscale
19.68% mIoU 19.99% 30.75%

Paper preset: t4_domain_d4 · from pmh.benchmark.presets import get_preset

Note: Notebook runs PMHTrainer(train_mode='feature_diff') + estimate_multilayer on demo loaders. Paper GTA5→Cityscapes: the paper (GTA5→Cityscapes).

Subtasks (paper)

Cityscapes rare-5 mIoU

E1_multiscale +11.1 pp mIoU.

Preset: t4_domain_d4

Build rare-5 training subset

Pixel-aligned per-layer TDI

Preset: t4_domain_d4

Run with matching-pmh

from pmh import PMHTrainer, evaluate_robust_fit
# nuisance="domain_shift"

Do not use PMH when

Single-layer hook enough — try T4A first.

Replace demo data with yours

Swap demo loaders for your train_loader, source_batches, target_batches, and deploy holdout. Hook the backbone before your task head.

← All 13 tasks · Quickstart