Implementation:Kornia Kornia IO Maps Configuration

Knowledge Sources	Kornia
Domains	ONNX, Pipeline_Design
Last Updated	2026-02-09 15:00 GMT

Overview

Pattern for configuring input-output tensor name mappings between chained ONNX models in Kornia's ONNXSequential.

Description

This is a configuration pattern, not a library API. The io_maps parameter of ONNXSequential is a list of (output_name, input_name) tuples that specify how to connect consecutive models. Users must inspect each model's graph to determine the correct tensor names. The mapping is applied when combining ONNX graphs into a single sequential pipeline.

Usage

Define io_maps when creating ONNXSequential with models whose output/input names do not automatically align.

Code Reference

Repository	https://github.com/kornia/kornia
File	`kornia/onnx/sequential.py`
Lines	L52–70
Interface	`io_maps: list[tuple[str, str]]` parameter for `ONNXSequential` constructor
Import	N/A (configuration parameter, not standalone import)

I/O Contract

Inputs

Parameter	Type	Description
`io_maps`	`list[tuple[str, str]]`	Defines `(output_name, input_name)` pairs between consecutive models

Outputs

Configured routing for the sequential pipeline.

Usage Examples

Mapping resize output to classifier input

from kornia.onnx import ONNXSequential

# Connect the "resized" output of the resize model
# to the "input_image" input of the classifier model
pipeline = ONNXSequential(
    "hf://operators/resize",
    "hf://operators/classifier",
    io_maps=[("resized", "input_image")],
)

Related Pages

Page Connections

Double-click a node to navigate. Hold to expand connections.

Principle

Implementation

Heuristic

Environment