image
Image feature types.
Classes:
| Name | Description |
|---|---|
ImageFeature |
Feature representing an image input or output. |
ImageFeature
#
Feature representing an image input or output.
This feature models image data as a multidimensional numerical tensor with pixel intensities arranged in channel-last format. It is typically used to represent visual inputs such as photographs, rendered scenes, or segmentation masks.
The expected tensor shape is (batch_size, height, width, num_channels),
where:
- height and width are the spatial dimensions of the image,
- num_channels corresponds to the color or modality channels (for example,
1 for grayscale, 3 for RGB, 4 for RGBA, or more for multispectral images).
Notes
- Image data must use the channel-last layout (H x W x C).
- The numerical type and normalization of pixel values depend on the associated preprocessing pipeline (for example, values in [0, 1] or [0, 255]).
- Current limitation: Xpdeep assumes fixed image dimensions across all instances of the dataset; resizing or cropping should be handled during preprocessing.
Methods:
| Name | Description |
|---|---|
to_model |
Convert to an |
from_model |
Image from model. |
Attributes:
| Name | Type | Description |
|---|---|---|
name |
Literal['IMAGE']
|
|
name: Literal['IMAGE'] = 'IMAGE'
#
to_model() -> ImageFeatureType
#
Convert to an ImageFeatureType instance.
Returns:
| Type | Description |
|---|---|
ImageFeatureType
|
Model-level representation of the image feature, including its static structural definition. |