Skip to content

pix specifications

volta2030 edited this page Dec 2, 2025 · 7 revisions

PIX is...

Pixelated Input eXtractor

Purpose

Integrating deep learning elements with images.

The future we pursue is training with only pix without any other configuration files.

Format 1.0

{
    "width": 4,
    "height": 4,
    "channel": 1,
    "depth": 8,
    "compression": {
        "method": "none"
    },
    "hex_data": "aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa",
    "label": {
        "cat": [[22, 44]],
        "dog": [[100, 103], [1, 1, 2, 2, 3, 3]],
    },
    "joint": {
    }
}

(0, 0) Coordinates

top-left corner

Clone this wiki locally