Machine learning engineer, Computer vision scientist, Data engineering enthusiast
python c++ c lua javascript java
git, spark, pandas, Tensorflow, sklearn, pandas, AWS, Docker Flask, SQL, ROS, OpenCV, WebGL, three.js, Android Studio MATLAB, Linux, vim
Using Machine Learnign to stream-line survey creation. Designed, built and deployed a multi-lingual model to production. Resulted in 3% subscription boost. Check it out on surveymonkey
more infoReal-Time floor segmentation on-board turtlebot robot with Jetson TX2. The segmentation is being done with a modified FCN8s network running on the Jetson.
Check it outRegular SLAM can localize the robot and create a map. But, the map that it creates is only a point cloud and carries no semantic meaning. Using deep learning, we can create the point cloud with semantic labels. Thanks to Jetson TX2, now, all of this can be done onboard and in real time.
Developed a single image 3D AR system with scale at Pair3D. The system is developed on an AWS and is accessible through a web API
Check it outDeveloped a 3D reconstruction API that accepts images or videos and generates and visualizes output models. Different methods are available to achieve the best performance for each scenario.
Check it outDeveloped relocalization using image matching. The video is showcasing our matching system performance. In this example, I used google street view API to get real images from streets. Two different positions of the camera were chosen each with 360-degree view. For each image in the first camera position (left image), the closest match to the second camera position (right image) was found using the system.
In this paper, we introduce a novel approach to implicitly utilize temporal data in videos for online semantic segmentation. The method relies on a fully convolutional network that is embedded into a gated recurrent architecture.
paperHelping robots to see things as humans do by teaching them.
paper