[GSoC 2026] Introduction & Interest in Project #1 — PerceptionMetrics | Vinit Jain

[vinitjain2005]

Hi @dpascualhe and @SakhinetiPraveena,

My name is Vinit Jain, a computer science student from India with a strong foundation in Python, Deep Learning, and PyTorch. I have successfully set up PerceptionMetrics locally on Windows, explored the codebase, and interacted with the GUI across all three tabs — Dataset Viewer, Inference, and Evaluator.

While exploring the Evaluator tab, I identified three input validation bugs that caused silent failures and confusing user experience for new users. I documented them in issue #436 and immediately submitted a PR with fixes — adding clear error messages, pre-run validation for the predictions output directory, and helpful hints explaining why the Run Evaluation button is disabled.

I am particularly excited about extending the GUI to support image and LiDAR segmentation visualization, and adding support for industry-standard datasets like SemanticKITTI and Cityscapes. I believe my hands-on experience with the codebase and my proactive approach to finding and fixing real bugs demonstrates my commitment to making PerceptionMetrics more robust and accessible for the wider community.

• Technical Demo: https://youtu.be/Y8MdV_D4lAk
• Issue #436:
  [Bug/Enhancement] GUI Evaluator tab silently fails and lacks pre-run input validation PerceptionMetrics#436
• PR:
  fix: add input validation and user-friendly error messages in evaluator tab PerceptionMetrics#437
• GitHub: https://github.com/vinitjain2005

Looking forward to contributing further and discussing my GSoC proposal!

Best regards,
Vinit Jain