Views
No views yet
1import coremltools as ct
2import numpy as np
3from PIL import Image
4
5# Load model
6model = ct.models.MLModel("dinov3_vitb16_448x448_int8.mlpackage")
7
8# Prepare image
9image = Image.open("image.jpg").resize((448, 448))
10
11# Extract features
12output = model.predict({"image": image})
13features = output["features"] # Shape: [1, embed_dim, grid_size, grid_size]1import CoreML
2
3// Load model
4guard let model = try? MLModel(contentsOf: modelURL) else {
5 fatalError("Failed to load model")
6}
7
8// Prepare image
9guard let image = UIImage(named: "image.jpg") else {
10 fatalError("Failed to load image")
11}
12
13// Extract features
14let input = try MLFeatureValue(image: image.cgImage!)
15let output = try model.prediction(from: [input])
16let features = output.featureValue(for: "features")?.multiArrayValue1@article{dinov3,
2 title={DINOv3: A Versatile Vision Foundation Model},
3 author={Meta AI Research},
4 journal={arXiv preprint arXiv:2508.10104},
5 year={2025}
6}
