withClassifierOptions static method
Creates a YOLO instance with classifier options for custom preprocessing
This constructor is specifically designed for classification models that need custom preprocessing, such as 1-channel grayscale models. 1-channel support is detected automatically from the model's input tensor.
Supported keys: labels (class names when the model has no embedded
metadata), enableColorInversion, enableMaxNormalization, inputMean,
and inputStd.
Example:
final yolo = YOLO.withClassifierOptions(
modelPath: 'assets/handwriting_model.tflite',
task: YOLOTask.classify,
classifierOptions: {
'enableColorInversion': true,
'enableMaxNormalization': true,
},
);
if need custom Normalization:
final grayscaleOptions = {
'enableMaxNormalization': false,
'inputMean': 127.5,
'inputStd': 127.5,
// 'labels': [...] (if the model has no embedded labels)
};
Implementation
static YOLO withClassifierOptions({
required String modelPath,
YOLOTask? task,
required Map<String, dynamic> classifierOptions,
bool useGpu = true,
bool useMultiInstance = false,
}) {
return YOLO(
modelPath: modelPath,
task: task,
useGpu: useGpu,
useMultiInstance: useMultiInstance,
classifierOptions: classifierOptions,
);
}