Semantic Segmentation (png images)

Export archive directory and file structure

|-- label_classes.json
|-- dataset1_name
    |-- image1.png
    |-- image2.png
    |-- ...
|-- dataset2_name
    |-- image1.png
    |-- image2.png
    |-- ...
|-- ...

label_classes.json structure

The file contains the definition of all label classes and corresponding pixel values of the images

    "png_index": 1,
    "class_name": "bedclothes",
    "color": "#4df3ce",
    "class_type": "object"
    "png_index": 2,
    "class_name": "building",
    "color": "#4c4ffc",
    "class_type": "object"

The Label Class object


png_index integer\
Pixel index

class_name string\
The name of the class

color string\
Associated with the label class color, in format #RRGGBBAA

class_type string Class type, "object" or "background"

Image description

All the images contain a single channel and are stored as 8-bit pixels, black and white. 0 - is a background.

Last updated on Jun 01, 2022

Removing the risk from vision AI.

Only 13% of vision AI projects make it to production, with Hasty we boost that number to 100%.