• Developed a real-time small image object detection and alert system based on C++ and Python. Applied OpenCV library for real-time trajectory detection with alert video generation, interruptible local and remote control, and high fault tolerance. Optimized the algorithm, reducing the processing time from 2 seconds to 40 milliseconds per frame.

• Collaborated with the hardware team, building the camera and lidar controller to scan and generate colorized point cloud using PCL and ROS. Improved the point cloud quality by adaptively controlling the scan duration and turning angle.

• Researched real-time CV-based detection system to detect human fallings using Python, Alphapose, Openpose, Pytorch, and Darknet. Compared the efficiency of methods and settled on openpose + ST-GCN method to detect human fallings in real-time.

• Experimented several Wide-angle lens undistortion algorithms to correct Mustache distortion of an image using OpenCV library.

• Designed and implemented model using OpenCV library to detect laser spot in the image. Analyzed performance of the model under different lighting conditions.

• Developed detection system for granary to detect bugs and to propose treatment plans using Pytorch, Darknet, and PaddlePaddle.

• Trained and evaluated deep learning models to detect the types and locations of the bugs from an image.

• Applied different data augmentation techniques to improve mAP of the model from 65% to 72%.

• Collaborated with a 3+ people team to build the model for recommendation system using Scikit-Learn, Tensorflow, R, and Hive.

• Built datasets from order history by randomly removing dishes from orders, and performed feature engineering to build user profiles.

• Applied data balancing and features selection to improve the accuracy of the model to around 70%.

• Held office hour to help students with questions related to the homework and projects.