Research

My research interests are in domain adaptation (via generative modeling, semi/self/un-supervised learning), multimodal learning and task-agnostic learning. In general, I am interested in solving problems that stem from a lack of annotated or domain-specific data and problems where testing data is distributionally different from training data.

In my current job, I primarily work on object detection, tracking and segmentation using camera data and point cloud data. I graduated from the University of Florida in 2016. Before that, I worked at the Indian Institute of Science under Prof. K R Ramakrishnan at the Computer Vision and Artificial Intelligence (CVAI) lab where I worked on touch-interface applications using a projector-camera based setup.

---

Patents

• A Perincherry, C Cruise “Domain Generation via Learned Partial Domain Translations” US Patent 11321587 - Generative modeling to generate novel domain data.

• A Perincherry, K Singh, N Nagraj Rao “Vehicle Intersection Operation” US Patent 11420625 - CNN+LSTM based model to predict vehicle right-of-way at intersections.

• N Nagraj Rao, A Perincherry “Sensor Domain Adaptation” US Patent App. 17/330692 - Generative modeling to translate legacy sensor data domain to newer domain.

• A Perincherry, A Mordovanakis, S Suthar, A Chand “Neural Network Object Identification” US Patent App. 17/228765 - Radar camera sensor fusion to perform object 3D shape identification.

• A Perincherry, I Patel, K Min “RCCC to RGB Domain Translation with Deep Neural Networks” US Patent 11068749 - Generative modeling to translate automotive sensor domain to common domains.

• N Jaipuria, G Sholingar, V Murali, R Bhasin, A Perincherry “Vehicle Image Generation” US Patent 11042758 - Generative modeling to improve driving safety features via domain adaptation.

---

Publications

• Adarsh Appaiah (equal contribution), Akhil Perincherry (equal contribution), Ajinkya Sanjeev Keskar and Vijaya Krishna “Spectrum Sensing in Cognitive Radio Based on Compressive Measurements”. International conference on Emerging trends in Communication, Control, Signal Processing and Computing Applications (IEEE-C2SPCA) Bangalore, India 2013

---

Knowledge Distillation Series (Paper Summaries)

• SimCLR
• SimCLRv2
• Big Self-Supervised Models Advance Medical Image Classification
• Contrastive Training for Improved Out-of-Distribution Detection
• Improving domain adaptation in de-identification of electronic health records through self-training
• Transformers
• ViT
• DALL-E
• Training data-efficient image transformers & distillation through attention
• BERT Pre-Training of Image Transformers
• Masked Autoencoders Are Scalable Vision Learners

Past Research

Rat whisker tracking - Univ. of Florida (2016)

---

Table-top touch interface using Kinect - Indian Institute of Science (2014)

---

Compressive sensing as a solution to Cognitive Radio (2013)

---

Other Projects

• Adaptive Speaker Recognition using Online Learning

  A novel technique was proposed to perform speaker recognition using adaptive algorithms such as Recursive Least Squares (RLS), Normalized Least Mean Squares (NLMS). The filter coefficients were seen to model the voice characteristics. The developed model was compared with the state of the art MFCC-VQ technique.

---

• Motor Imagery Classification (2015) Motor imagery tasks are tasks where a person imagines performing a task and the neural signals are recorded. The model classifies between left finger and tongue imagery movements using Common Spatial Pattern and Support Vector Machines from EEG recorded signals. The model developed recorded an accuracy of 92%, the highest that has been reported for the BCI 2000 dataset at the time.

---

• Manatee Call Detection Using Recursive Least Squares (2015)

  The problem of detecting Manatee calls in a signal plus noise situation is studied. Trained models were obtained for Manatee calls and noise, and an ensemble of models were applied and compared.

---

• Majorized Multi-class Kernel SVM

  A Majorized Multi-class kernel SVM was derived and designed from scratch, and compared with other ML algorithms such as Random Forests, Logistic Regression, Deep Neural Networks, Decision Trees and libSVM implementation.

---

Outreach

• Cluster coordinator for South-East Bangalore, Youth For Seva (2013-2014) - Involved in sapling planting and providing kids with computer training.

• MentorUF, University of Florida (2016) - Mentored a middle-school student for a year.

• IEEE Eta Kappa Nu, University of Florida (2016)

Teaching

• Teaching Assistant for graduate course “Foundations of Digital Signal Processing” at University of Florida during fall 2015.