Registry / ai-ml / imx500-converter

imx500-converter

library3.18.2pypypi✓ verified 83d ago

Official Sony Semiconductor Solutions converter for the IMX500 intelligent vision sensor. Version 3.18.2 converts AI models (TensorFlow, PyTorch, ONNX) to the IMX500 firmware format. Released monthly.

ai-ml devops

pip install imx500-converter

INSTALL

IMPORT

SIG · IMX500-CONVERTER

imx500-converter

ai-mlpythonv3.18.2

harness data pending

Install & Compatibility

Where this runs

No compatibility data collected yet for this library.

Code

Verified usage

Verified import paths — ran on the pinned version, not inferred.

converter

✓ from imx500_converter import converter

Underscore in package name; dash is used on PyPI but module uses underscore.

Converts an ONNX model to IMX500 format. Adjust input_shape to match your model.

from imx500_converter import converter result = converter.convert( model_path='model.onnx', output_path='output.sdf', input_shape=(1, 3, 224, 224), output_compression=0 ) print('Conversion success:', result['status'])

Debug

Known issues

breakingIn version 3.0+, the module name changed from 'imx500-converter' to 'imx500_converter'. Import statements from older versions will break.

fix

Replace 'import imx500_converter' or 'from imx500_converter import ...' accordingly.

affects: >=3.0.0

gotchaModel input shapes are in NCHW format (batch, channels, height, width) by default. Using NHWC will cause tensor mismatches.

fix

Ensure input_shape matches the model's expected layout. For TensorFlow models, use NHWC and transpose if needed.

affects: all

deprecatedThe 'output_compression' parameter in v3.0 defaults to 0 but may be removed in future versions. Compression levels above 0 are not recommended for accuracy-critical models.

fix

Set output_compression=0, or omit it. For high accuracy, use post-training quantization instead.

affects: >=3.0.0, <4.0.0

Errors

Common errors & fixes

ModuleNotFoundError: No module named 'imx500_converter'

The package is installed as 'imx500-converter', but the import uses underscores. Also missing dependency if using pip from local wheel.

fix

Run 'pip install imx500-converter' and then import with 'from imx500_converter import converter'.

TypeError: convert() got an unexpected keyword argument 'input_shape'

The API changed between v2 and v3. In v2, the argument was called 'input_dims' or similar.

fix

Use 'input_shape' instead of 'input_dims'. Check the documentation for the exact parameter name.

ValueError: Input shape must be a tuple of 4 integers

Passed a list or wrong number of dimensions.

fix

Provide input_shape as a tuple of (batch, channels, height, width), e.g., (1, 3, 224, 224).

Upgrade

Version history

3.18.2latest on PyPI · released Apr 22, 2026

Audit

Dependencies

No dependency data recorded yet.

Agent activity

11 hits · last 30 days

node

Bingbot

Resources

No resource links recorded.