1.18.3

• Improved stability of the overall model transformation when the input tensor contains two or more undefined dimensions.
• Resize
  • Added optimization process for Resize with undefined dimensions.
  • Valid only if the input tensor is fixed in NHWC.
  • Axis transposition correction by JSON in this pattern is no longer necessary.
  • https://github.com/PINTO0309/onnx2tf/files/12758261/retinaface_onnx_dynamic.onnx.zip
    
  • convert

    onnx2tf \
    -i retinaface_onnx_dynamic.onnx \
    -osd \
    -cotof

    onnx	Before tflite	After tflite

  
• Cannot use converted model with dynamic input shape #521

What's Changed

• Added optimization process for Resize with undefined dimensions by @PINTO0309 in #523

Full Changelog: 1.18.2...1.18.3