mlboydaisuke/EfficientAD-CoreML

EfficientAD (MVTec bottle) — Core ML

WACV 2024

Visual anomaly detection for industrial inspection. Outputs a per-pixel anomaly heatmap plus a scalar score from a 256x256 image, so defects are localised rather than just flagged.

PDN-Small teacher/student pair plus an autoencoder: the anomaly map is the disagreement between teacher-student and autoencoder-student. This checkpoint is trained on the MVTec AD bottle category — retrain per category for other objects.

Core ML conversion of openvinotoolkit/anomalib for on-device inference on iPhone, iPad and Mac. Converted with coremltools; the packages are stateless, so all sequencing and buffering lives in your Swift code.

Task	image segmentation
Upstream	openvinotoolkit/anomalib
Packages	1
Download size	14 MB
Minimum iOS	17.0

Files

File	Size	Compute units	SHA-256
EfficientAD_Bottle.mlpackage.zip	14 MB	all	-
Total	14 MB

compute_units is not a suggestion -- it is the configuration the conversion was verified against. Moving a package to a different compute unit can silently change the numerics (FP16 attention overflow) or crash on the GPU.

Download

hf download mlboydaisuke/coreml-zoo --include "efficientad/*" --local-dir ./efficientad
unzip './efficientad/efficientad/*.zip' -d ./efficientad

Use in Swift

import CoreML

let config = MLModelConfiguration()
config.computeUnits = .all   // as converted — see the table above

// Unzip the .mlpackage, drop it into your Xcode target and Xcode compiles it
// at build time:
let model = try EfficientAD_Bottle(configuration: config)

// ...or compile a downloaded .mlpackage at runtime:
let compiled = try await MLModel.compileModel(at: mlpackageURL)
let model = try MLModel(contentsOf: compiled, configuration: config)

Demo

• Sample app — sample_apps/EfficientADDemo, a standalone SwiftUI project.
• Models Zoo — this model is downloadable and runnable inside the Models Zoo app on the App Store, no build required.

Conversion

• Script: convert_efficientad.py
• Pitfalls hit during conversion (FP16 overflow, ANE buffer limits, stride handling): docs/coreml_conversion_notes.md
• Model index: CoreML-Models

License

The conversion inherits the upstream license: MIT.

Credits

• Upstream authors: openvinotoolkit/anomalib, 2023
• Core ML conversion: john-rocky (Daisuke Majima)