Nikhil Makkar

I'm currently a founding engineer at Auric AI Labs in Bangalore. Before Auric I was at Purdue, where I worked with Dr. Daniel Aliaga (Computer Science), Dr. Jinha Jung (Civil Engineering), and Dr. Soowon Chang (School of Construction) on separate research problems. Earlier, I was in the GeoAI group at Oak Ridge National Laboratory, working with Dr. Lexie Yang and Dr. Dalton Lunga, and I also did my MS (Electrical Engineering) at the University of Houston with Dr. Saurabh Prasad.

Email/Scholar/Substack/LinkedIn

Research

Interested in Computer Vision and Multi-modal Learning.

Updates

[Aug 2026] Benchmarked SOTA geo foundation models (GFMs) on high-res SAR images and wrote about it. Post
[Mar 2026] Tried to trace the research around Reasoning from CoT to geospatial models. Post
[Jan 2026] My mental model around research progress in geospatial foundation models. Post
[Apr 2025] Joined Auric AI Labs as the first employee and founding engineer.
[Feb 2025] Moved back to India.
[Dec 2024] Left Purdue with an MS in Civil Engineering (Dropped out of PhD).
[Mar 2024] Active learning work from ORNL finally accepted at J-STARS (phew!).

Publications

Cross-sensor building extraction results

Boundary-aware Adversarial Learning Domain Adaption and Active Learning for Cross-sensor Building Extraction

H. L. Yang, N. Makkar, M. Laverdiere, A. Rose

IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 2024

Active learning (+ domain adaptation) pipeline that matches full supervision performance on a fraction of the annotation.

paper/

@article{yang2024boundary,
  title   = {Boundary-aware adversarial learning domain adaption and active
             learning for cross-sensor building extraction},
  author  = {Yang, Hsiuhan Lexie and Makkar, Nikhil and
             Laverdiere, Melanie and Rose, Amy},
  journal = {IEEE Journal of Selected Topics in Applied Earth Observations
             and Remote Sensing},
  year    = {2024}
}

Forest inventory plot center measurement from UAS imagery

Improving Deciduous Forest Inventory Plot Center Measurement using Unoccupied Aerial Systems Imagery

J. Carpenter, D. Rentauskas, N. Makkar, J. Jung, S. Fei

Journal of Forestry, 2023

UAV data collection of Martell forest. Imagery locates inventory plot centers more accurately than the GPS methods field crews normally rely on.

paper/

@article{carpenter2023improving,
  title   = {Improving deciduous forest inventory plot center measurement
             using unoccupied aerial systems imagery},
  author  = {Carpenter, Joshua and Rentauskas, Daniel and Makkar, Nikhil and
             Jung, Jinha and Fei, Songlin},
  journal = {Journal of Forestry},
  volume  = {121},
  number  = {3},
  pages   = {262--271},
  year    = {2023}
}

Discriminative domain adaptation results

Adversarial Learning based Discriminative Domain Adaptation for Geospatial Image Analysis

N. Makkar, H. L. Yang, S. Prasad

IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 2021

Discriminative adversarial alignment moves a classifier to a new sensor with no labels in the target domain.

paper/

@article{makkar2021adversarial,
  title   = {Adversarial learning based discriminative domain adaptation
             for geospatial image analysis},
  author  = {Makkar, Nikhil and Yang, Hsiuhan Lexie and Prasad, Saurabh},
  journal = {IEEE Journal of Selected Topics in Applied Earth Observations
             and Remote Sensing},
  year    = {2021}
}

Learning to Count Grave Sites for Cemetery Observation Models with Satellite Imagery

D. Lunga, R. Dhamdhere, S. Walters, L. Bragg, N. Makkar, M. Urban

IEEE Geoscience and Remote Sensing Letters, 2020

Hybrid workflow of weak localization plus regression model for detecting dense and small objects for supporting population modeling.

paper/

@article{lunga2020learning,
  title   = {Learning to count grave sites for cemetery observation models
             with satellite imagery},
  author  = {Lunga, Dalton and Dhamdhere, Rohit and Walters, Sarah and
             Bragg, Lexi and Makkar, Nikhil and Urban, Marie},
  journal = {IEEE Geoscience and Remote Sensing Letters},
  year    = {2020}
}

Entropy and boundary based adaptation results

Entropy and Boundary based Adversarial Learning for Large Scale Unsupervised Domain Adaptation

N. Makkar, H. L. Yang

IEEE International Geoscience and Remote Sensing Symposium (IGARSS), 2020

Penalizing prediction entropy and boundary disagreement helps in adversarial adaptation.

paper/

@inproceedings{makkar2020entropy,
  title     = {Entropy and boundary based adversarial learning for large
               scale unsupervised domain adaptation},
  author    = {Makkar, Nikhil and Yang, Hsiuhan Lexie},
  booktitle = {IEEE International Geoscience and Remote Sensing Symposium
               (IGARSS)},
  pages     = {589--592},
  year      = {2020}
}

Large Scale Unsupervised Domain Adaptation of Segmentation Networks with Adversarial Learning

X. Deng, H. L. Yang, N. Makkar, D. Lunga

IEEE International Geoscience and Remote Sensing Symposium (IGARSS), 2019

Output feature adaptation of segmentation network for roads and building extraction.

paper/

@inproceedings{deng2019large,
  title     = {Large scale unsupervised domain adaptation of segmentation
               networks with adversarial learning},
  author    = {Deng, Xueqing and Yang, Hsiuhan Lexie and Makkar, Nikhil and
               Lunga, Dalton},
  booktitle = {IEEE International Geoscience and Remote Sensing Symposium
               (IGARSS)},
  pages     = {4955--4958},
  year      = {2019}
}

Small/Fun Projects