Image Semantic Segmentation based on the state-of-art DeepLab Tensorflow model.
The Semantic Segmentation
is the process of associating each pixel of an image with a class label, (such as flower, person, road, sky, ocean, or car).
Unlike the Instance Segmentation
, which produces instance-aware region masks, the Semantic Segmentation
produces class-aware masks.
For implementing Instance Segmentation
consult the Object Detection Service instead.
The Semantic Segmentation Processor
uses the Semantic Segmentation Function library and the TensorFlow Service.
The incoming type is byte[]
, and the content type is application/octet-stream
. The processor processes the input byte[]
image and outputs augmented byte[]
image payload and json header.
Processor’s input is an image byte array, and the output is an augmented image byte array, and a JSON header semantic_segmentation
in this format:
[
[ 0, 0, 0 ],
[ 127, 127, 127 ],
[ 255, 255, 255 ],
...
]
The output header json format represents the color pixel map computed from the input image.
- semantic.segmentation.color-map-uri
-
Every pre-trained model is based on certain object color maps. The pre-defined options are: - classpath:/colormap/citymap_colormap.json - classpath:/colormap/ade20k_colormap.json - classpath:/colormap/black_white_colormap.json - classpath:/colormap/mapillary_colormap.json (String, default:
classpath:/colormap/citymap_colormap.json
) - semantic.segmentation.debug-output
-
save output image inn the local debugOutputPath path. (Boolean, default:
false
) - semantic.segmentation.debug-output-path
-
<documentation missing> (String, default:
semantic-segmentation-result.png
) - semantic.segmentation.mask-transparency
-
The alpha color of the computed segmentation mask image. (Float, default:
0.45
) - semantic.segmentation.model
-
pre-trained tensorflow semantic segmentation model. (String, default:
https://download.tensorflow.org/models/deeplabv3_mnv2_cityscapes_train_2018_02_05.tar.gz#frozen_inference_graph.pb
) - semantic.segmentation.output-type
-
Specifies the output image type. You can return either the input image with the computed mask overlay, or the mask alone. (OutputType, default:
<none>
, possible values:blended
,mask
)