DEDL Introduces novel Autoencoder post-processing for image segmentation which improves upon traditional segmentation architecture by 3% at no extra space and inference time. We also introduce data-efficient deep learning methods for medical image classification and segmentation which improves upon previous approaches by 26% and 5%. Arixv Paper , Papers with Code , Accepted at ECCV - Medical Computer Vision Workshop 2022
Co-authored a paper on “Melanoma Classification using efficient nets with multiple ensembles and patient-level data”, International Conference on Computational Intelligence - ICCI 2020, IIIT Pune.
Top 57% in Lyft's Motion Prediction for Autonomous Vehicles.
Top 43% in Google Research's Open Images Object Detection RVC 2020 edition.