pydlmeta 1.0.3

No project description provided

PyDLmeta

Features: identify the model type belong to which deep learning framework and extract the meta data if possible. The meta data includes input/output name and shape of the model.

The supported formats are:

• Tensorflow: frozen model *.pb, *.h5, SavedModel directory, *.tflite

• Pytorch: *.pt, TorchScript

• ONNX: *.onnx

• Caffe model directory: *.caffemodel/ *.prototxt

• Openvino IR directory: *.xml/ *.bin

Installation

• Create a Python 3.8 environment and activate it.

git clone --depth 1 -b develop --recursive https://github.com/skymizer/pydlmeta.git
(cd pydlmeta && python3 -m pip install -e .)

Usage

— Retrieve the metadata of the model

from pydlmeta.meta import retrieve_model_metadata
res = retrieve_model_metadata("/path/to/your/model")

• Identify model format

from pydlmeta.identifier.model import identify
model_format = identify(model_path)