Added Aliked support for onnx export

[rohankhaire-work]

This Pull request references the issue #69. I have added onnx export for aliked.

[manhtd20z]

Do you think aliked, superpoint or sift, which is better extractor ?

[ahojnnes]

This doesn't work for dynamic input image sizes for me. I get errors like these when passing images with different input dimension:

E:onnxruntime:, sequential_executor.cc:572 ExecuteKernel] Non-zero status code returned while running Reshape node. Name:'/block3/conv1/Reshape_5' Status Message: /private/tmp/onnxruntime-20250819-46528-71tomy/onnxruntime/core/providers/cpu/tensor/reshape_helper.h:47 onnxruntime::ReshapeHelper::ReshapeHelper(const TensorShape &, TensorShapeVector &, bool) input_shape_size == size was false. The input tensor cannot be reshaped to the requested shape. Input shape:{1,32,66,66}, requested shape:{1,1,32,50,66}

[manhtd20z]

This doesn't work for dynamic input image sizes for me. I get errors like these when passing images with different input dimension:

E:onnxruntime:, sequential_executor.cc:572 ExecuteKernel] Non-zero status code returned while running Reshape node. Name:'/block3/conv1/Reshape_5' Status Message: /private/tmp/onnxruntime-20250819-46528-71tomy/onnxruntime/core/providers/cpu/tensor/reshape_helper.h:47 onnxruntime::ReshapeHelper::ReshapeHelper(const TensorShape &, TensorShapeVector &, bool) input_shape_size == size was false. The input tensor cannot be reshaped to the requested shape. Input shape:{1,32,66,66}, requested shape:{1,1,32,50,66}

it's because the shape converted is fixed with number of keypoint output. seem like we need add padding static

[ahojnnes]

@rohankhaire-work @manhtd20z Any idea how to solve the issue?

[manhtd20z]

just using threshold is -1. The error will solve.

[ahojnnes]

@manhtd20z Your suggested fix does not work for me, assuming you mean the max_num_keypoints as the "threshold"?