Bashitha P.

I convert your vision models from a notebook executable file to a production ready application. The real reason you cannot run your custom vision models with real-time accuracy and latency is the un-optimized tech stack and inconsistent training data. What I Build and Deliver: ✅ Dataset annotation and analysis (OpenCV, Roboflow) ✅ Object detection & segmentation | YOLO, RT-DETR, SAM ✅ ByteTrack, BoT-SORT based multi-object tracking ✅ Pose estimation & motion analysis using YOLO-pose, MediaPipe ✅ Application specific feature extraction (e.g. polygon zone based counting) ✅ Model optimization for high speed inference through OpenVINO, TensorRT, ONNX, and FP16/INT8 quantization ✅ Edge deployment on Jetson Orin, Raspberry Pi, Mobile phones ✅ LLM agent integration for alert automation and reasoning ✅ Real-time inference (10 - 30 FPS) and (50 - 200 ms) latency - depends on the hardware stack I am an Engineer with 2+ years hands-on experience in deploying Machine Vision pipelines under heavy real-world lighting and occlusions, which are the key practical issues most engineers don't solve. Domain Expertise: ➡️ Intersection traffic motion analysis (delay times, vehicle density/detection zone) ➡️ Product anomaly detection, production line tracking, off-site reading of analog gauges, worker safety monitoring ➡️ Ball tracking for trajectory analysis, real-time pit lane monitoring ➡️ Theft detection, shelf monitoring at retail stores ➡️ Fabric edge detection for automated sewing, depth based grasp estimation with 3D cloud generation ⚙️ Technical Stack: Python, C++ | OpenCV | Pytorch, TensorFlow | YOLOv5, YOLOv8, YOLOv11, YOLO26 | DeepSORT, ByteTrack, BoT-SORT | RT-DETR, RF-DETR | MediaPipe, OpenPose | FastAPI, Docker, AWS SageMaker | ONNX, OpenVino, TensorRT, CoreML, NCNN | Roboflow, CVAT, LabelImg | NVIDIA Jetson, Raspberry Pi 5, Coral TPU | LangChain, Claude, Gemini, GPT 📩 Let's talk if you have a project or an idea that involves visual data. I will layout my technical approach and we can discuss any alterations before you commit to the project.