ImageClassificationPerceptor Objects

class ImageClassificationPerceptor(MultiPlatformPerceptorBase)

ImageClassificationPerceptor is a class that implements the Perceptor interface for image classification.


def __init__(processor_preference: dict,
             threshold: float,
             top_k: int = None,
             mean: float = 128.0,
             std: float = 128.0,
             quantized: bool = True,
             num_cpu_threads: int = 1)


  • processor_preference: A dictionary of processor preference. The key is the processor. Values are dictionaries of model paths and labels.
  • threshold (float): The threshold for object detection.
  • top_k (int): The number of top predictions to return.
  • mean (float): The mean of the image (Coral Edge TPU).
  • std (float): The standard deviation of the image (Coral Edge TPU).
  • quantized (bool): Whether the model is quantized (CPU).
  • num_cpu_threads (int): The number of threads to use for inference (CPU). Defaults to 1.


from darcyai.perceptor.image_classification_perceptor import ImageClassificationPerceptor
from darcyai.perceptor.processor import Processor
processor_preference = {
    Processor.CORAL_EDGE_TPU: {
        "model_path": "/path/to/model.tflite",
        "labels_file": "/path/to/labels.txt", // The path to the labels file.
    Processor.CPU: {
        "model_path": "/path/to/model.tflite",
        "labels": { // A dictionary of labels.
            "label_1": "label_1_name",
            "label_2": "label_2_name",
image_classification_perceptor = ImageClassificationPerceptor(
    processor_preference=processor_preference, threshold=0.5, top_k=5)