Yen Nhi Ta

- Developed and optimized real-time AI CCTV systems for industrial safety inspection and Vehicle Management Systems (VMS) with RTSP cameras; implemented multi-threaded processing for multiple cameras, optimized streaming performance, and improved inference time, accuracy. Successfully deployed in large-scale factories. - Deployed and optimized AI environments on Linux servers and edge devices (Jetson, IPC), achieving up to 30% faster real-time inference speed. - Executed AI projects for product quality inspection, including visual defect detection and production line product counting. Hands-on experience with industrial cameras (Hikvision, Basler, etc.). Applied image generation techniques to create synthetic data, significantly reducing annotation time for OCR and visual defect detection tasks. - Create a CI/CD pipeline and build a Docker image to automate the deployment process in the development and production environments.

Developed Colony Counter machine, utilizing Deep Learning and Image Processing for colony recognition algorithm. Accomplish over 90% accurate identification, differentiate colonies by color, and precisely measure size. Program and package GUI using WPF with features similar to the video at https://www.youtube.com/watch?v=i-LaRL8o2sg Designed lighting system for optimal performance.

𝗥𝗲𝘀𝗲𝗮𝗿𝗰𝗵 𝗦𝘁𝘂𝗱𝗲𝗻𝘁 - Research, develop and complete enterprise project (𝗗𝗲𝘀𝗶𝗴𝗻 𝗮𝗻𝗱 𝗣𝗿𝗼𝗴𝗿𝗮𝗺𝗺𝗶𝗻𝗴 𝗼𝗳 𝗜𝗺𝗮𝗴𝗲 𝗣𝗿𝗼𝗰𝗲𝘀𝘀𝗶𝗻𝗴 𝗳𝗼𝗿 𝗕𝗮𝗰𝘁𝗲𝗿𝗶𝗮 𝗰𝗼𝘂𝗻𝘁𝗶𝗻𝗴 𝗮𝗻𝗱 𝗰𝗹𝗮𝘀𝘀𝗶𝗳𝗶𝗰𝗮𝘁𝗶𝗼𝗻 𝗳𝗼𝗿 𝘁𝗵𝗲 𝗦𝗰𝗮𝗻 𝗺𝗮𝗰𝗵𝗶𝗻𝗲). - Research trainning techniques: Deep Learning (Object Detection, Segmentation, Classification). - Mini projects: Detect and classify traffic signs use Yolov4 (accuracy: 91%), Unlock recognition using facial recognition and hand gestures, Vehicle license plate detection. 𝗧𝗲𝗮𝗰𝗵𝗲𝗿 𝗔𝘀𝘀𝗶𝘀𝘁𝗮𝗻𝘁 - Supporting teacher in preparing teaching programs related to Image Processing subjects.

𝗠𝗮𝗶𝗻 𝗿𝗲𝘀𝗽𝗼𝗻𝘀𝗶𝗯𝗶𝗹𝗶𝘁𝗶𝗲𝘀: - Responsibility for developing the Image Processing projects with the AI team, and participating in some projects like detecting defective objects, counting objects, and OCR - OCV, which decreased 20% total time in manufacturing processing. - Collecting, processing, and augmenting data for the training program. - Programming GUI (Qt Designer, .NET) for mini tasks of the project