kitcar_ml.traffic_sign_detection.fasterrcnn package

Submodules

kitcar_ml.traffic_sign_detection.fasterrcnn.evaluate module

kitcar_ml.traffic_sign_detection.fasterrcnn.export module

kitcar_ml.traffic_sign_detection.fasterrcnn.inference module

kitcar_ml.traffic_sign_detection.fasterrcnn.model module

Classes:

Model(class_names[, pretrained])

class Model(class_names: List[str], pretrained: bool = True)[source]

Bases: DetectionModel

Methods:

`predict`(images, **kwargs)	Take in a list of images and predict the bounding boxes.
`fit`(data_loader, val_data_loader[, epochs, ...])	Train the model on the given data_loader.
`save`(file)	Save the internal model weights to a file.
`load`(file)	Load a model from a .pth file containing the model weights.
`export_to_onnx`(output_file)	Export this model into a onnx format.
`half`()	Switch to FP16 values instead of FP32 to speed up inference.

Attributes:

`DEFAULT_OPTIMIZER_KWARGS`
`_abc_impl`

__get_prediction_indices(scores: List[float], min_score=0, max_iou=0.2)

Apply non maximum suppression and min_score.

Returns resulting indices.

predict(images: List[ndarray], **kwargs) → List[Tuple[List[ndarray], List[str], List[float]]][source]

Take in a list of images and predict the bounding boxes.

Returns: A list of bounding boxes with labels and scores.

DEFAULT_OPTIMIZER_KWARGS = {'lr': 0.005, 'momentum': 0.9, 'weight_decay': 0.0005}

fit(data_loader: DataLoader, val_data_loader: DataLoader, epochs: int = 10, optimizer_name: str = 'SGD', optimizer_args: Dict[str, float] = {'lr': 0.005, 'momentum': 0.9, 'weight_decay': 0.0005}, visualize: bool = False, tensorboard_path: str = 'runs') → List[float][source]

Train the model on the given data_loader.

If given a validation data_loader, returns a list of loss scores at each epoch.

save(file: str)[source]

Save the internal model weights to a file.

Parameters: file – The name of the file. Should have a .pth file extension.

classmethod load(file: str) → Model[source]

Load a model from a .pth file containing the model weights.

Parameters: file – The path to the .pth file containing the saved model.
Returns: The model loaded from the file.

export_to_onnx(output_file: str)[source]

Export this model into a onnx format.

Parameters: output_file – Path to the output file

half()[source]

Switch to FP16 values instead of FP32 to speed up inference.

This does only work on CUDA.

_abc_impl = <_abc_data object>