Akash Karthikeyan

Hello There!

Drop a message here!!

Email  /  CV  /  kaggle  /  Github  /  Google Scholar  /  LinkedIn

profile photo
AvatarOne: Monocular 3D Human Animation
Akash Karthikeyan, Robert Ren, Yash Kant, Igor Gilitschenski

WACV, 2024
Reconstructing realistic human avatars only from monocular videos
CAMM: Building Category-agnostic and Animatable 3D Models from Monocular Videos
Tianshu Kuai, Akash Karthikeyan, Yash Kant, Ashkan Mirzaei, Igor Gilitschenski

CVPRW, 2023
Kinematic aware, Animatable and Class-agnostic 3D deformable objects from causual monocular videos without the use any priors or camera poses - A Template free approach
Lyft L5 toolkit - Motion forcasting
Kaggle - Build motion prediction models for self-driving vehicles | 2021

Motion prediction with the help of rasterized set of images, CNN , LSTM encoder-decoder based model.
Also planning to explore the kalman and other optimisation techniques other than learning based approaches
Also, currenly exploring a Vectornet based approach skipping the whole razterization part and making the pipeline faster.
Segment and label helmets in video footage
Kaggle - NFL, 2021

• Detector to find helmets, Image2Map (BEV)
• Classifier to classify players into 2(H/V) teams and Registration of detected players on 2D map to provided tracking data. Later track detected
bounding boxes and reassign players.
• Implementation of Kalman for improvised tracking alongside testing of DeepSORT
• Predict the 2022 College Men’s Basketball Tournament
• Analyse the trend based on past 5 year’s data

Automated Annotation and Classification of Catheters in Chest X-Rays
Akash Karthikeyan, Saravana Perumaal Subramanian,
ICCCSP, 2022

Annotate and classify cathater position. Robust approach to get the endpoints of the cathaters even though the enpoints contribute to a less than 1% footprint of the whole image

Google Deciemter Challenge Kaggle | Google, 2022
Results and Architecture

• Improve high precision GNSS positioning and navigation accuracy on smartphones
• Process and clean the GNSS logs to compute location down to decimeter or even centimeter resolution placed 94/810
• Use of Kalman smoothening and constant velocity heading model to improve accuracy of GNSS data, more visualization in repo

Action Recognition and Tracking of People in Realtime

End2End detection, pose-estimation and tracking with ReID of crowded regions under various lighting conditions

Landmark Recognition | Google

ICCV 2022 Workshop Competition | 58/930

• Use of Additive Angular Margin Loss (ArcFace), and other Bag-of-tricks from person re-identification
• Random Erasing, label smooth, triplet loss, IBN-extension, last stride=1
• Used DELF and DOLG based approches to find and extract features/ Similar stuff used here also for Happy Dolphin comp

Bag-Vave-Mask Ventilator

• Design and prototype a slider-crank based actuation mechanism to automate the compression in BVM ventilator
• Performed Pressure trajectory analysis to achieve required PIP and PEEP values at outlet with the help of COMSOL multiphysics module

Zero Energy Air-Cooler
Winner | TNSI - 2022 (certificate)

• Device fabrication and characterization of MOF as advanced moisture sorbents for energy-efficient high temperature cooling
• Performed XRD and Thermogravimetric Analysis to determine the structure of MOF and water update respectively.

More Projects

  • MMDet Ultimate Detection and Segmentation Toolkit: I experimented with MMDet Pipeline for medical image segmentation. code here Based on competition hosted here,
    Placed 119/1505 participants
  • G2Net Gravitational Wave Detection - European Gravitational Observatory: Find gravitational wave signals from binary black hole collisions Used FFT and 1D CNN to detect the gravitational waves.
    Placed 33/1219 teams
  • March Machine Learning Mania 2022 - Men’s: Predict the 2022 College Men's Basketball Tournament (kaggle competition), also here is the visulalization tool for bracket builing NCAA_BracketBuilder
    Placed 14/930 participants
  • Crypto- Market prediction based on anonymised features (Ubiquant Market Prediction) I Used Tabnet based model.
    Placed 99/2893 participants

TEAM YUKTA RACING - Autobots Assemble!
We Build, We Race.....! Defending Champions of GKDC 2020

I borrowed this template from Jon Barron's website.