[Technical Questions] How to deal with a problem when integrating ML models

[Turquoise-T]

In what area do you have a technical challenge?

Other (non-StanfordBDHG-related) area

Description

When I tried to integrate an ML model into my code to recognize images, I encountered an issue:

"This neural network model does not have a parameter for requested key 'precisionRecallCurves'. Note: only updatable neural network models can provide parameter values, and these values are only accessible in the context of an MLUpdateTask completion or progress handler."

I couldn't find an answer, and I'm not sure if anyone knows how to handle this issue. I did not pass or request the precisionRecallCurves parameter. It seems that the problem might be related to whether the model is updatable.

Here is the relevant part of my code:

@MainActor
class ImageClassifier: ObservableObject {
    // MARK: - Published Properties
    @Published var image: UIImage?
    @Published var classificationOptions: [String] = []
    @Published var isProcessing: Bool = false
    @Published var errorMessage: String?

    // MARK: - Internal Properties
    private let confidenceThreshold: Float = 0.7
    private var cancellables = Set<AnyCancellable>()

    // MARK: - Initialization
    init() {
        setupBindings()
    }

    private func setupBindings() {
        $image
            .dropFirst()
            .sink { [weak self] newImage in
                if let image = newImage {
                    self?.classifyImage(image)
                }
            }
            .store(in: &cancellables)
    }

    // MARK: - Main Classification Method
    func classifyImage(_ image: UIImage) {
        DispatchQueue.main.async {
            self.isProcessing = true
            self.errorMessage = nil
        }

        guard let ciImage = CIImage(image: image) else {
            self.handleError("Could not create CIImage from UIImage")
            return
        }

        guard let model = try? VNCoreMLModel(for: SeeFood(configuration: MLModelConfiguration()).model) else {
            self.handleError("Failed to load SeeFood model")
            return
        }

        let request = VNCoreMLRequest(model: model) { [weak self] request, error in
            guard let self = self else { return }

            if let error = error {
                self.handleError("Classification failed: \(error.localizedDescription)")
                return
            }

            guard let results = request.results as? [VNClassificationObservation] else {
                self.handleError("Could not process classification results")
                return
            }

            let validResults = results.filter { $0.confidence >= self.confidenceThreshold }
            let topResults = validResults.prefix(3)

            DispatchQueue.main.async {
                self.classificationOptions = topResults.map {
                    "\($0.identifier) (\(String(format: "%.1f", $0.confidence * 100))%)"
                }
                self.isProcessing = false
            }
        }

        let handler = VNImageRequestHandler(ciImage: ciImage)
        DispatchQueue.global(qos: .userInitiated).async {
            do {
                try handler.perform([request])
            } catch {
                print("Failed to perform classification: \(error.localizedDescription)")
            }
        }
    }

    // MARK: - Helper Methods
    private func handleError(_ message: String) {
        DispatchQueue.main.async {
            self.errorMessage = message
            self.isProcessing = false
        }
    }
}```
The error occurs in the following block:
```swift
let handler = VNImageRequestHandler(ciImage: ciImage)
DispatchQueue.global(qos: .userInitiated).async {
    do {
        try handler.perform([request])
    } catch {
        print("Failed to perform classification: \(error.localizedDescription)")
    }
}```

### Reproduction

None

### Expected behavior

have expected output for the model.(I tried MobileNetV2 and seeFood)

### Additional context

_No response_

### Code of Conduct

- [X] I agree to follow this course's Code of Conduct and Contributing Guidelines

[PSchmiedmayer]

@felixschlegel Do you mind taking a look at this?

[felixschlegel]

Can you please provide a repo + branch for running this code?

[Turquoise-T]

Hi,
This problem was solved on Tuesday's class. Thanks your all for the kind help!