CSPDarkNet backbones

• Original Link : https://keras.io/api/keras_cv/models/backbones/csp_darknet/
• Last Checked at : 2024-11-25

source

CSPDarkNetBackbone class

keras_cv.models.CSPDarkNetBackbone(
    stackwise_channels,
    stackwise_depth,
    include_rescaling,
    use_depthwise=False,
    input_shape=(None, None, 3),
    input_tensor=None,
    **kwargs
)

This class represents the CSPDarkNet architecture.

Reference

• YoloV4 Paper
  • CSPNet Paper
  • YoloX Paper

For transfer learning use cases, make sure to read the guide to transfer learning & fine-tuning.

Arguments

• stackwise_channels: A list of ints, the number of channels for each dark level in the model.
• stackwise_depth: A list of ints, the depth for each dark level in the model.
• include_rescaling: bool, whether to rescale the inputs. If set to True, inputs will be passed through a Rescaling(1/255.0) layer.
• use_depthwise: bool, whether a DarknetConvBlockDepthwise should be used over a DarknetConvBlock, defaults to False.
• input_shape: optional shape tuple, defaults to (None, None, 3).
• input_tensor: optional Keras tensor (i.e. output of keras.layers.Input()) to use as image input for the model.

Returns

A keras.Model instance.

Examples

input_data = tf.ones(shape=(8, 224, 224, 3))
# Pretrained backbone
model = keras_cv.models.CSPDarkNetBackbone.from_preset(
    "csp_darknet_tiny_imagenet"
)
output = model(input_data)
# Randomly initialized backbone with a custom config
model = keras_cv.models.CSPDarkNetBackbone(
    stackwise_channels=[128, 256, 512, 1024],
    stackwise_depth=[3, 9, 9, 3],
    include_rescaling=False,
)
output = model(input_data)

source

from_preset method

CSPDarkNetBackbone.from_preset()

Instantiate CSPDarkNetBackbone model from preset config and weights.

Arguments

• preset: string. Must be one of “csp_darknet_tiny”, “csp_darknet_s”, “csp_darknet_m”, “csp_darknet_l”, “csp_darknet_xl”, “csp_darknet_tiny_imagenet”, “csp_darknet_l_imagenet”. If looking for a preset with pretrained weights, choose one of “csp_darknet_tiny_imagenet”, “csp_darknet_l_imagenet”.
• load_weights: Whether to load pre-trained weights into model. Defaults to None, which follows whether the preset has pretrained weights available.

Examples

# Load architecture and weights from preset
model = keras_cv.models.CSPDarkNetBackbone.from_preset(
    "csp_darknet_tiny_imagenet",
)
# Load randomly initialized model from preset architecture with weights
model = keras_cv.models.CSPDarkNetBackbone.from_preset(
    "csp_darknet_tiny_imagenet",
    load_weights=False,

source

CSPDarkNetTinyBackbone class

keras_cv.models.CSPDarkNetTinyBackbone(
    stackwise_channels,
    stackwise_depth,
    include_rescaling,
    use_depthwise=False,
    input_shape=(None, None, 3),
    input_tensor=None,
    **kwargs
)

CSPDarkNetBackbone model with [48, 96, 192, 384] channels and [1, 3, 3, 1] depths.

Reference

• YoloV4 Paper
  • CSPNet Paper
  • YoloX Paper

For transfer learning use cases, make sure to read the guide to transfer learning & fine-tuning.

Arguments

• include_rescaling: bool, whether or not to rescale the inputs. If set to True, inputs will be passed through a Rescaling(1/255.0) layer.
• input_tensor: optional Keras tensor (i.e. output of layers.Input()) to use as image input for the model.
• input_shape: optional shape tuple, defaults to (None, None, 3).

Example

input_data = tf.ones(shape=(8, 224, 224, 3))
# Randomly initialized backbone
model = CSPDarkNetTinyBackbone()
output = model(input_data)

source

CSPDarkNetSBackbone class

keras_cv.models.CSPDarkNetSBackbone(
    stackwise_channels,
    stackwise_depth,
    include_rescaling,
    use_depthwise=False,
    input_shape=(None, None, 3),
    input_tensor=None,
    **kwargs
)

CSPDarkNetBackbone model with [64, 128, 256, 512] channels and [1, 3, 3, 1] depths.

Reference

• YoloV4 Paper
  • CSPNet Paper
  • YoloX Paper

For transfer learning use cases, make sure to read the guide to transfer learning & fine-tuning.

Arguments

• include_rescaling: bool, whether or not to rescale the inputs. If set to True, inputs will be passed through a Rescaling(1/255.0) layer.
• input_tensor: optional Keras tensor (i.e. output of layers.Input()) to use as image input for the model.
• input_shape: optional shape tuple, defaults to (None, None, 3).

Example

input_data = tf.ones(shape=(8, 224, 224, 3))
# Randomly initialized backbone
model = CSPDarkNetSBackbone()
output = model(input_data)

source

CSPDarkNetMBackbone class

keras_cv.models.CSPDarkNetMBackbone(
    stackwise_channels,
    stackwise_depth,
    include_rescaling,
    use_depthwise=False,
    input_shape=(None, None, 3),
    input_tensor=None,
    **kwargs
)

CSPDarkNetBackbone model with [96, 192, 384, 768] channels and [2, 6, 6, 2] depths.

Reference

• YoloV4 Paper
  • CSPNet Paper
  • YoloX Paper

For transfer learning use cases, make sure to read the guide to transfer learning & fine-tuning.

Arguments

• include_rescaling: bool, whether or not to rescale the inputs. If set to True, inputs will be passed through a Rescaling(1/255.0) layer.
• input_tensor: optional Keras tensor (i.e. output of layers.Input()) to use as image input for the model.
• input_shape: optional shape tuple, defaults to (None, None, 3).

Example

input_data = tf.ones(shape=(8, 224, 224, 3))
# Randomly initialized backbone
model = CSPDarkNetMBackbone()
output = model(input_data)

source

CSPDarkNetLBackbone class

keras_cv.models.CSPDarkNetLBackbone(
    stackwise_channels,
    stackwise_depth,
    include_rescaling,
    use_depthwise=False,
    input_shape=(None, None, 3),
    input_tensor=None,
    **kwargs
)

CSPDarkNetBackbone model with [128, 256, 512, 1024] channels and [3, 9, 9, 3] depths.

Reference

• YoloV4 Paper
  • CSPNet Paper
  • YoloX Paper

For transfer learning use cases, make sure to read the guide to transfer learning & fine-tuning.

Arguments

• include_rescaling: bool, whether or not to rescale the inputs. If set to True, inputs will be passed through a Rescaling(1/255.0) layer.
• input_tensor: optional Keras tensor (i.e. output of layers.Input()) to use as image input for the model.
• input_shape: optional shape tuple, defaults to (None, None, 3).

Example

input_data = tf.ones(shape=(8, 224, 224, 3))
# Randomly initialized backbone
model = CSPDarkNetLBackbone()
output = model(input_data)

source

CSPDarkNetXLBackbone class

keras_cv.models.CSPDarkNetXLBackbone(
    stackwise_channels,
    stackwise_depth,
    include_rescaling,
    use_depthwise=False,
    input_shape=(None, None, 3),
    input_tensor=None,
    **kwargs
)

CSPDarkNetBackbone model with [170, 340, 680, 1360] channels and [4, 12, 12, 4] depths.

Reference

• YoloV4 Paper
  • CSPNet Paper
  • YoloX Paper

For transfer learning use cases, make sure to read the guide to transfer learning & fine-tuning.

Arguments

• include_rescaling: bool, whether or not to rescale the inputs. If set to True, inputs will be passed through a Rescaling(1/255.0) layer.
• input_tensor: optional Keras tensor (i.e. output of layers.Input()) to use as image input for the model.
• input_shape: optional shape tuple, defaults to (None, None, 3).

Example

input_data = tf.ones(shape=(8, 224, 224, 3))
# Randomly initialized backbone
model = CSPDarkNetXLBackbone()
output = model(input_data)