Tools to convert ONNX files (NCHW) to TensorFlow format (NHWC)

• Fixed the number of NMS outputs to be fixed outputs in max_output_boxes_per_class. -onwdt option to generate a dynamic tensor. Taking into account that tflite's primary use is for Android, change the behavior so that it is replaced with a fixed output shape by default.

  -onwdt, --output_nms_with_dynamic_tensor
    The number of bounding boxes in the NMS output results is
    not fixed at the maximum number of max_output_boxes_per_class,
    but rather at the smallest possible number of dynamic tensors.
    If this option is disabled, NMS output is padded to the number
    set in the max_output_boxes_per_class attribute.
    e.g.
    disable --output_nms_with_dynamic_tensor:
        output_tensor_shape: [100, 7]
    enable --output_nms_with_dynamic_tensor:
        output_tensor_shape: [N, 7]
  

  • disable -onwdt

  • enable -onwdt

Yolov7-tiny to TensorflowLite comversion results in a dynamic output model incompatible with TfLite Java API #159

• To support GPU Delegate, replace all Gather processing part of NMS with strided_slice by @PINTO0309 in https://github.com/PINTO0309/onnx2tf/pull/162
• Fixed the number of NMS outputs to be fixed outputs in max_output_boxes_per_class. -onwdt option to generate a dynamic tensor. by @PINTO0309 in https://github.com/PINTO0309/onnx2tf/pull/164

Full Changelog: https://github.com/PINTO0309/onnx2tf/compare/1.5.37...1.5.38