Onnx2tf

- `DepthToSpace`
- Fix incorrect conversion of `DepthToSpace` in `CRD` mode
- `common_functions`
- Consider tuple as input shape for `pre_explicit_broadcast`
- `Flatten`
- Use Keras `Flatten()` if there are two or more undefined dimensions
- `Slice`
- Consider the case where `starts`, `ends`, and `steps` are variables
- `IF`
- Consider the case where a constant is included in the output of a subgraph

What's Changed
* Fix incorrect conversion of DepthToSpace in CRD mode by Shakarim94 in https://github.com/PINTO0309/onnx2tf/pull/92
* Fix `pre explicit broadcast`, `Flatten`, `Slice`, `IF` by PINTO0309 in https://github.com/PINTO0309/onnx2tf/pull/94

New Contributors
* Shakarim94 made their first contribution in https://github.com/PINTO0309/onnx2tf/pull/92

**Full Changelog**: https://github.com/PINTO0309/onnx2tf/compare/1.3.14...1.3.15

1. Fix `explicit_broadcast` to consider scalar values
python
If shape is empty (scalar value), return it without doing anything.
if graph_node_input_shape1 == [] or graph_node_input_shape2 == []:
return const_or_var_1, const_or_var_2

2. [WIP] Experimental implementation of `Loop`

What's Changed
* Fix explicit_broadcast to consider scalar values, and experimental implementation of `Loop` by PINTO0309 in https://github.com/PINTO0309/onnx2tf/pull/90


**Full Changelog**: https://github.com/PINTO0309/onnx2tf/compare/1.3.13...1.3.14

- `ConvTranspose`
- If the spartial shape is an undefined dimension, run inference with dummy input tensors in onnxruntime to try to estimate the output shape.
- https://github.com/onnx/models/tree/main/vision/object_detection_segmentation/mask-rcnn

What's Changed
* If the spartial shape is an undefined dimension, run inference with dummy input tensors in onnxruntime to try to estimate the output shape. by PINTO0309 in https://github.com/PINTO0309/onnx2tf/pull/89


**Full Changelog**: https://github.com/PINTO0309/onnx2tf/compare/1.3.12...1.3.13

- Support for `Faster_R-CNN`
- https://github.com/PINTO0309/onnx2tf/releases/download/1.1.28/faster_rcnn-10.onnx
- `Flatten`
- Improved handling of undefined dimensions
- `ScatterElements`
- Improved handling of undefined dimensions
- Bug fixes
- `Scatter`
- Disable debug information output
- `TopK`
- MLIR only accepts scalar values for k-values, thus compressing the dimension
- `Unsqueeze`
- Improved handling of undefined dimensions. Substitution by `ExpandDims`.

What's Changed
* Improved handling of undefined dimensions and bug fixes by PINTO0309 in https://github.com/PINTO0309/onnx2tf/pull/88


**Full Changelog**: https://github.com/PINTO0309/onnx2tf/compare/1.3.11...1.3.12

What's Changed
* Bug fixes `OneHot` by PINTO0309 in https://github.com/PINTO0309/onnx2tf/pull/85
* Bug fixes for `Inverse` and improved NHWC determination for `Resize` by PINTO0309 in https://github.com/PINTO0309/onnx2tf/pull/86
* Fixed to disable NHWC flag when transposition occurs in `pre_process_transpose` or `post_process_transpose` by PINTO0309 in https://github.com/PINTO0309/onnx2tf/pull/87


**Full Changelog**: https://github.com/PINTO0309/onnx2tf/compare/1.3.10...1.3.11

1. Improved handling of undefined dimensions
- `Clip`, `Concat`, `Transpose`, `common_functions`
2. Support for `Scatter` (ScatterElements)

What's Changed
* Improved handling of undefined dimensions by PINTO0309 in https://github.com/PINTO0309/onnx2tf/pull/84


**Full Changelog**: https://github.com/PINTO0309/onnx2tf/compare/1.3.9...1.3.10