Update RANSAC Voting for PyTorch (>1.10) & CUDA (>11.1) Compatibility

[Divyam10]

ℹ️ Ensure Compatibility with Newer PyTorch and CUDA Versions for RANSAC Voting Layer

🚀 Summary: This commit updates the RANSAC voting module to ensure compatibility with the latest PyTorch and CUDA versions.

🔄 Changes Made:

• Pytorch has removed Thc/thc.h headers from version >11.1 which gives compilation errors. Commented extern THCState* state; in ransac_voting.cpp
• Replaced the torch.solve() with torch.linalg.solve() for newer version.
• Applied tensor.ge() function on the mask as tensor.masked_select expects a boolean tensor array in newer versions.

📋 Additional Notes:

• Requirements.txt should be updated to work with newer versions as well.

🔧 Dependencies:

• PyTorch: >1.11
• CUDA: >11.1

🧪 Testing:

• Ran comprehensive tests to validate the compatibility with PyTorch [2.1.2] and CUDA [11.8].
• Verified that the RANSAC voting module performs as expected.

👏 Acknowledgments: Thanks to the authors for sharing code for this valuable research.

📅 Date: 21st Jan, 2024