Feat/efficientad light - #170

🔗 Related Issue

Fixes #

📝 Description

This PR adds EfficientAD support to AnomaVision Production Autopilot and extends the performance regression tests to cover all supported anomaly detection algorithms.

Main changes

• Added EfficientAD as an Autopilot candidate alongside PaDiM and PatchCore.

• Added --efficientad_model to the Autopilot CLI.

• Updated Autopilot model selection to compare PaDiM, PatchCore, and EfficientAD using the same validation data.

• Added EfficientAD latency and accuracy metrics to the Autopilot report.

• Preserved the existing fancy HTML production report and extended it to include EfficientAD.

• Added configuration-driven inference performance tests for all three algorithms.

• Updated the README with EfficientAD and Autopilot documentation.

Example:

anomavision autopilot \
  --config config.yml \
  --padim_model ./distributions/padim/bottle/anomav_exp/model.pt \
  --patchcore_model ./distributions/patchcore/bottle/anomav_exp/model.pt \
  --efficientad_model ./distributions/efficientad/bottle/anomav_exp/model.pt \
  --device cpu \
  --validation_split 1.0 \
  --target_latency_ms 50 \
  --output_dir ./production_package

🔄 Type of Change

• 🐛 Bug fix (non-breaking change which fixes an issue)

• 🚀 New feature (non-breaking change which adds functionality)

• 💥 Breaking change (fix or feature that would cause existing functionality to not work as expected)

• 📖 Documentation update

• 🏗️ Infrastructure / CI/CD update

🧪 Hardware & Matrix Testing

I have successfully built and tested this code using uv on:

• anomavision[cpu] (Standard/Edge)

• anomavision[cu121] (CUDA 12.1)

• anomavision[cu124] (CUDA 12.4)

• anomavision[cu118] (CUDA 11.8)

Host OS used for testing:

• Linux / Ubuntu

• Windows (Native or WSL2)

• macOS

Performance benchmark

The inference performance tests were executed for:

• PaDiM

• PatchCore

• EfficientAD

Example measured results on CPU:

The EfficientAD regression test correctly fails when the configured latency threshold is exceeded.

✅ Developer Checklist

• My code follows the core style guidelines of this project (Ruff/Black formatting).

• I have run the relevant pytest tests locally.

• Lockfile Guard: If I added or modified a dependency in pyproject.toml, I have run uv lock --python 3.10 and committed the updated uv.lock file.

• I have added tests that prove my fix is effective or that my feature works.

• I have updated the documentation accordingly.

📸 Screenshots / Visual Proof

The Autopilot HTML report includes:

• Model comparison cards

• Selected production model

• Image AUROC

• Pixel AUROC

• Median and P95 latency

• Anomaly coverage

• False-positive rate

• Calibrated threshold

• Localization health

• Deployment recommendation

The report is generated as:

production_package/production_autopilot_report.html