Akshay Sethi is a Machine Learning Engineer based in Toronto, Canada. He was previously at IBM Research (IBM-IRL, Bangalore) and graduated from IIIT-Delhi working primarily on deep learning, computer vision and biometrics.

Resume | CV | Google Scholar | Github | LinkedIn

• Survey Paper on Relational Representation Learning for Dynamic Graphs accepted to JMLR
• Received Best BTP (Thesis) Award and Best Academic Performance Award in the class of 2018, IIIT-D.
• Paper accepted in Pattern Recognition Letters on Facial Attribute Analysis.
• Paper accepted at AAAI 2018 on Automated Deep learning Papers to Code Converison.

System and Method for Guided Policy Generation from Data for Augmentation
Akshay Sethi, Srikant Tamilselvam, Anush Sankaran, Senthil Mani
In Filing

System and Method for Data Insights based Test Case Generation
Shreya Khare, Srikant Tamilselvam, Anush Sankaran, Senthil Mani, Akshay Sethi
In Submission

Relational Representation Learning for Dynamic Graphs: A Survey
Seyed Mehran Kazemi, Rishab Goel, Akshay Sethi, Pascal Poupart
Journal of Machine Learning Research (JMLR) [pdf]

Survey Paper on Dynamic Knowledge Graphs.

Coverage Testing of Deep Learning Models using Dataset Characterization
Akshay Sethi, Senthil Mani, Srikant Tamilselvam, Anush Sankaran
Under Submission to FSE [pdf]

We show that Classification accuracy is not enough to rank models for prediction in real world.

Multi-label Sparse Representation based Classification
Akshay Sethi, Angshul Majumdar, Mayank Vatsa, Richa Singh
Under Submission to TKDE

We modify the dictionary based sparse representation classifier to give multi-label outputs and get SOTA results on many datasets for this problem.

Deep Learning for Attribute Prediction and Small Sample Size Problems
Akshay Sethi, Richa Singh, Mayank Vatsa
Undergraduate Thesis, 2018 [pdf]

Use GANs for enlarging datasets and predict facial attributes using autoencoders.

Residual Codean Autoencoder for Facial Attribute Analysis
Akshay Sethi, Maneet Singh, Richa Singh, Mayank Vatsa
Pattern Recognition Letters, 2018 [pdf]

We propose residual autoencoder with custom loss for improved facial attribute prediction.

DLPaper2Code: Auto-generation of Code from Deep Learning Research Papers
Akshay Sethi, Anush Sankaran, Naveen Panwar, Shreya Khare, Senthil Mani
AAAI Conference on Artificial Intelligence (AAAI), 2018 [pdf]

Automated generation of Deep learning model code in popular deep learning libraries from the research papers that describe them.

DARVIZ: A Visually IDE to build Deep Learning Models
Anush Sankaran, Naveen Panwar, Shreya Khare, Senthil Mani, Akshay Sethi, Rahul Aralikatte, Neelamadhav Gantayat
AAAI Conference on Artificial Intelligence (AAAI) Demo Track, 2018 [pdf]

Demo paper for automated code extraction from Deep learning research papers.

DARVIZ: A Visual IDE to build Deep Learning Models
Shreya Khare, Naveen Panwar, Akshay Sethi, Anush Sankaran, Senthil Mani, Rahul Aralikatte, Neelamadhav Gantayat
CoDS-COMAD Demo Track, 2018 [pdf]

Demo paper for automated code extraction from Deep learning research papers.

Deep Neural Networks for Segmentation of Basal Ganglia Sub-Structures in Brain MR Images
Akshay Sethi, Akshat Sinha, Ayush Agarwal, Chetan Arora, Anubha Gupta
The Indian Conference on Computer Vision, Graphics and Image Processing (ICVGIP), 2016 [pdf]

We use Regularized Autoencoders for predicting voxels in Brain MRI scans which belong to Basal Ganglia Region.

Template: this