Skip to content

README.md

Latest commit

 

History

History
42 lines (31 loc) · 1.3 KB

README.md

File metadata and controls

42 lines (31 loc) · 1.3 KB
Raw

AF3AD Integrations

This directory contains detector-specific reference integrations. The reusable AF3AD synthesis code lives in the top-level af3ad package, while each integration keeps the training code and heavier detector dependencies.

Install The Integration Directory

The integration modules are imported from the AF3AD repository tree. There is no separate package installer for integrations; make the repository root importable before running a launcher:

cd /path/to/AF3AD
export AF3AD_ROOT="$PWD"
export PYTHONPATH="$AF3AD_ROOT:$PYTHONPATH"

Quick PO3AD import check after installing the PO3AD environment:

python - <<'PY'
from af3ad import PseudoAnomalySynthesizer
import integrations.po3ad

print("AF3AD PO3AD integration is importable")
PY

Environment Guides

  • po3ad/: PO3AD-style online training. This path needs Python 3.8, PyTorch 1.9, CUDA-compatible packages, and MinkowskiEngine built with OpenBLAS for legacy reproduction. It also includes a modern PyTorch 2.x
    • CUDA 12.x venv for testing checkpoints on newer GPUs.

Run the root launchers from the AF3AD repository root after activating the matching detector environment:

python scripts/train_po3ad_integration.py --dataset Real3D --dataset-base-dir data/Real3D --category airplane