Pannaga Sudarshan's Projects
Official Code: Implicit 3D Orientation Learning for 6D Object Detection from RGB Images
A query system for a hypothetical bank scenario
Code for "PVNet: Pixel-wise Voting Network for 6DoF Pose Estimation" CVPR 2019 oral
This is a simple weekend open-cv project. It uses concepts like masking, thresholding, detecting contours, calculating euclidian distances, drawing shapes over live video stream, adding texts, etc.
Comparitive analysis of decision tree and random forest models for predicting CVD
"DenseFusion: 6D Object Pose Estimation by Iterative Dense Fusion" code repository
An ETL logic is written in Spark for transforming the given data set present in S3, and query on the transformed data is run using AWS Redshift. The data sets are in json format. All the raw data in json format has to be first uploaded to an S3 source bucket. Using EMR, a Spark job is executed, which would fetch the source data from S3 source bucket, and then perform the necessary transformations on it as per the problem statement. Finally, store the transformed data were to partitioned and stored in parquet format in S3 destination bucket. Now, these files are accessed using AWS Redshift by running SQL queries on the transformed processed data.
The project eliminates the heuristic method of generating the university's timetable. The time table was developed in JAVA programming language and GUI was also developed. It also had personal preferences that were given to the faculties based on their seniority. Constraints were developed and assigned according to the real university norms. Ultimately time table was generated pertaining to a section, which was also printed and had an option to save as PDF.
iiwa pose estimation
Created point clouds from epipolar geometry along with PC correspondences using ICP. Later implemented PointNet from scratch to classify and predict the generated point clouds.
PyTorch implementation of the PoseCNN framework
(ECCV 22 Oral)Vote from the Center: 6 DoF Pose Estimation in RGB-D Images by Radial Keypoint Voting
RNNPose: Recurrent 6-DoF Object Pose Refinement with Robust Correspondence Field Estimation and Pose Optimization, CVPR 2022
Gen AI based 3D printing.