首先导入头文件
import CoreML
//图像识别框架。
import Vision
功能实现
func startIdentifyingPictures() {
//拖入素材，拿到图片URL
let path = Bundle.main.path(forResource: "111", ofType: "png")
let imageUrl = NSURL.fileURL(withPath: path!)
//拿到模型文件
let fileModel = Resnet50()//GoogLeNetPlaces()
// 把模型拿來做视觉处理
let model = try! VNCoreMLModel(for: fileModel.model)
//创建识别图片的请求头
let handler = VNImageRequestHandler(url: imageUrl)
//创建请求,completionHandler请求回调
let request = VNCoreMLRequest(model: model) { (request, error) in
//拿到识别结果
//1、判别结果是否存在
guard let results = request.results as? [VNClassificationObservation] else {
fatalError("识别结果为空")
}
//什么东西
var bestPrediction = ""
//相似度
var bestConfidence: VNConfidence = 0
//for遍历results
for classfication in results {
if classfication.confidence > bestConfidence {
bestPrediction = classfication.identifier
bestConfidence = classfication.confidence
}
//print("預測結果\(classfication.identifier) 可信度\(classfication.confidence)")
}
print("最終預測結果\(bestPrediction) 可信度\(bestConfidence)")
}
try! handler.perform([request])
}
