Aryan Mehesare
Back to projects
•
1 min read
Single Image Super Resolution (SRCNN)
  • Developed and trained a 3-layer SRCNN in PyTorch on 800+ high-res images (DIV2K), achieving strong PSNR/SSIM scores.
  • Compared model performance across 2×, 3×, and 4× scaling factors using MLflow with 50+ experiment runs.
  • Managed training infrastructure on AWS SageMaker with S3-backed data pipelines for 20GB+ of training data.