| 20,000+ Fresh Resumes Monthly | |
|
|
| | Click here or scroll down to respond to this candidateEXPERIENCESKILL HIGHLIGHTSEDUCATIONCONTACTCandidate's Name
Street Address Oakland StreetPhiladelphia, PA, USAPHONE NUMBER AVAILABLEEMAIL AVAILABLEEMAIL AVAILABLEBachelor of Science:Computer Engineering 2013University of Illinois Urbana-Champaign, ILAlgorithmMathC++PythonComputer VisionDeep LearningLSTMCNNACHIEVEMENT19th place in ACM/ICPCWorld Finals 2010AI & Algorithm ArchitectApr 2022 To Jun 2024Aug 2019 To Nov 2021meTech1021meTech1021Library Media Database Indexing and Retrieval System AI & Algorithm EngineerData Collection: Gathered a diverse set of labeled images and videos retrieval system based on each category.Preprocessing: Applied resizing, normalization, and data augmentation to clean and prepare the images and videos for analysis. Extract key frames from videos for uniform processing.Feature Extraction: Implemented a convolutional neural network(CNN) for image data and a combination of CNN and LSTM for video data to extract relevant features from images and temporal features from video sequences.Database Construction: Organize extracted features and metadata(labels, timestamps) into a structured database, which supports efficient querying and retrieval.Indexing and Search: Developed a retrieval system using techniques nearest neighbor search to allow for fast and accurate matching of user queries with content in the database.3D Photosphere Generation AlgorithmAlgorithm DeveloperKeypoint Extraction: Utilized feature detection algorithm in Python and C++ to extract keypoints from each photo, leveraging GPU acceleration with CUDA for improved performance.Keypoint Matching: Implemented feature matching algorithms to match keypoints between adjacent photos. Optimized the matching process using parallel processing techniques on the GPU to handle large datasets efficiently.Homography Estimation: Estimated the homography matrix or camera pose between each pair of matched photos using RANSAC(Random Sample Consensus) to filter outliers, ensuring robust performance in various lighting conditions and scenes.Image Stitching: Warped and stitched the images together based on the computed homographies. Leveraged CUDA to accelerate the image transformation and blending processes.Cloud Infrastructure: Deployed the entire workflow on AWS, utilizing auto-scaling to manage resource allocation dynamically based on workload demands, ensuring efficient processing of large batches of images. |