Skip to content
View MayeshMohapatra's full-sized avatar

Highlights

  • Pro

Block or report MayeshMohapatra

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Maximum 250 characters. Please don't include any personal information such as legal names or email addresses. Markdown supported. This note will be visible to only you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
MayeshMohapatra/README.md

Hi there, I'm Mayesh Mohapatra πŸ‘‹

Master of Computer Science student at Texas A&M University focusing on the intersection of 3D Computer Vision, Deep Learning, and Biomechanical Analysis.

Current research at TAMU involves developing modular 3D reconstruction pipelines for dynamic sports analysis, specifically investigating the decoupling of camera ego-motion from global human trajectories.


πŸ”­ Research & Current Focus

  • Graduate Researcher at TAMU: Developing pipelines to transform 2D footage into 3D actionable data for biomechanical analysis.
  • 3D Vision Specialization: Working on multi-view multi-person reconstruction, addressing temporal inconsistency and "foot sliding" artifacts in architectures like TRAM and WHAM.
  • Deep Learning Reproducibility: Implementing state-of-the-art research papers from scratch (e.g., Recurrent Attention Models, custom ResNet/VGG variants).

πŸ› οΈ Technical Toolbox

Category Tools & Technologies
Languages Python, C/C++, SQL (Postgres, MySQL), R
Deep Learning PyTorch, TensorFlow, CUDA Toolkit, PyTorch Profiler
3D Vision SLAHMR, WHAM, Human3R, SAM3D, VGGT
Scientific Computing NumPy, Pandas, Scikit-Learn, Matplotlib, Seaborn
Spatial & Imagery GDAL, Rasterio, Cartopy, OpenCV
Data & Cloud Spark, PySpark, Dask, AWS, Docker, Weights & Biases

πŸ“‚ Featured Research Projects

πŸŒ€ [CycloAI]

End-to-end ML pipeline for cyclone detection and intensity estimation.

  • Developed an automated pipeline processing ~14,000 satellite images with ISRO Space Applications Centre.
  • Reduced inference latency from hours to <1 second through model pruning and quantization.
  • Impact: Co-authored peer-reviewed work accepted at IEEE IGARSS 2025.

πŸ€– [Recurrent Attention Model (RAM)]

Reproducing the RAM architecture in PyTorch using reinforcement learning.

  • Achieved 96.64% test accuracy using only ~49% of image pixels (6 glimpses).
  • Implemented via policy-gradient reinforcement learning (REINFORCE).

πŸ›‘οΈ [GAN-based Insider Threat Detection]

Research interned at Georgia Institute of Technology.

  • Improved model robustness by 15% using GANs to generate synthetic insider-threat datasets.

🀝 Connect with Me

LinkedIn


Pinned Loading

  1. ChestXRayClassification ChestXRayClassification Public

    Implementation of DenseNet-121 on the CheXpert Dataset.

    Jupyter Notebook

  2. DiseaseDetectioninAppleLeaves DiseaseDetectioninAppleLeaves Public

    Jupyter Notebook 1

  3. ImageSegmentation-UNet ImageSegmentation-UNet Public

    Python

  4. CNNs-from-scratch CNNs-from-scratch Public

    A collection of State of the Art Convolutional Neural Network based architectures which are implemented from the ground up.

    Jupyter Notebook