Animesh Gupta

Animesh Gupta



I’m a Final-year Electronics and Computer Engineering undergraduate student at Thapar University, Patiala, India. My research focus is broadly centered around Computer Vision and Deep Learning. I work on research projects dealing with Generative Adversarial Networks, Knowledge Distillation, and zero-shot classification. I intend to explore the neural networks under the low-resource training data scenario using Knowledge Distillation and the Efficient Subset selection paradigm.

I am currently working as a Research Intern at UiT Norway. In the past, I have also worked in NVIDIA and SketchX as a Research Intern. I have also worked as a part-time Research Engineer at a startup, Minus Zero, where I worked on an autonomous electric car with level 5 autonomy.


  • [July 2022] Adaptive Fine-Grained Sketch-Based Image Retrieval is accpeted in the ECCV 2022.
  • [May 2022] Joined UiT Norway as a Research Intern.
  • [March 2022] Joined NVIDIA as a Research Intern.
  • [July 2021] Joined SketchX lab as a Research Intern.
  • [June 2021] Recieved Grant by Weights & Biases for ML Reproducibility Challenge, Spring 2021.
  • [May 2021] Volunteered for the ICLR Conference 2021.
  • [September 2020] Won Bronze Metal for Kaggle Notebook in I’m Something of a Painter Myself challenge.
  • [July 2020] Top 42% Worldwide in Google Landmark Recognition 2020.


  • Knowledge Distillation
  • Computer Vision
  • Deep Learning


  • Bachelor's in Electronics and Computer Engineering, 2023

    Thapar University



Research Intern

UiT Norway

May 2022 – Present Remote

Comparing the performance of ResNet and ViT models when trained using the subset of the dataset with different percentages of entire dataset (CIFAR10, CIFAR100, Medical Imaging dataset) (1%, 5%, 10%, and so on).

  • Subsets are created in a manner such that training model results in performance with minimal loss when compared to entire dataset.
  • Examining the effects of using the pretrained weights and random initialized models on the subsets of training data.

Research Intern


Mar 2022 – May 2025 Remote

Experimented with latest Real-Time Lane Detection work and vision transformers for an improved solution for DRIVE-Perceptron platform with faster inference and performance. Responsibilities include:

  • Done literature survey and tested recent solution on Real-Time Lane Detection.

Research Intern


Jul 2021 – Mar 2022 Remote

Worked on Fine-Grained Sketch Based Image Retrieval and Category-Level Sketch Based Image Retrieval. Responsibilities include:

  • Developed techniques for Datafree Category Level Sketch Based Image Retrieval.
  • Used Explainable AI techniques to identify the importance of particular strokes in the sketch images.
  • Extracted 3D models and 2D images by scraping websites for new datasets of shoes and chairs.
  • Contributed to one of the papers which created an adaptive Fine-Grained Sketch-Based Image Retrieval model. It adapts to new categories or different sketching patterns at test time, published in ECCV 2022.


GirlScript Summer of Code

Mar 2021 – Jun 2022 Remote

Worked on adding architectural examples and demos to different Open Source repository. Responsibilities include:

  • Face-X: Added NasNet and Xception model architecture for Face Recognition.
  • Comet.Box: Added YOLOv5 example for the object detection.

Research Engineer (Part-time)

Minus Zero

Nov 2020 – Mar 2021 Patiala

Worked on the Road Segmentation problem for autonomous cars in India. Responsibilities include:

  • Used FCHarDNet as base architecture and trained on the Indian driving dataset (10k images and 34 classes).
  • Modelling

Research Intern

Indian Institute of Information Technology Allahabad

Nov 2020 – Jan 2021 Allahabad

Worked on scene text detection problem with state of the art model. Responsibilities include:

  • Testing various kind of neural networks
  • Modelling


ZSl Generative

An Open Source Zero Shot classification toolbox based on PyTorch.

I’m Something of a Painter Myself

Developed a GAN that generates 7,000 to 10,000 Monet-style images.

AI for Blind

Developed a classification model for predicting seven emotions like (angry, disgusted,fearful, happy, neutral, sad and surprised) using FER-2013 dataset.

Google Landmark Recognition 2020

Developed a classification model for predicting landmark labels using the GLDv2 dataset.

Open Source Contributions


  • Added new figures in python for the Kevin Murphy’s book “Probabilistic Machine Learning: An Introduction”.


  • To make new geo-locations accessible to new mappers added several new presets.


  • Added improvements (like modals, dark mode bugs) for enhancing the use of GUI interface.


  • Added NasNet and Xception model architecture for the face recognition.


  • Added YOLOv5 example for the object detection.

d2l study group

  • Maintainer of the study group with daily discussions with the students of our college on the book Dive into deep learning.

DSC (Thapar University OfficialWebsite)

  • Improved repository readability for new user navigation.


  • Virtual conference toolkit. Added video links and issue tracker bots.