tensorrt-optimization

YOLOE-Unified is a novel framework that integrates YOLOE with distilled CLIP, runtime SAM refinement, and TensorRT optimization for efficient open-vocabulary object detection and instance segmentation on edge devices (Jetson Orin, etc.).

computer-vision zero-shot-learning instance-segmentation multimodal-fusion edge-deployment open-vocabulary-detection real-time-inference tensorrt-optimization

• Updated Jan 2, 2026
• Python

High-performance, multi-GPU computer vision pipeline for Genetec streams using NVIDIA DeepStream, Triton, and Numba JIT to track personnel, evaluate weapon threats, and perform in-VRAM OCR analytics.

object-tracking gstreamer-pipeline license-plate-recognition threat-detection ada-lovelace edge-ai triton-inference-server paddleocr pyds numba-jit nvidia-deepstream rtx-5090 nvidia-blackwell tensorrt-optimization mqtt-telemetry rtx-6000-ada cuda-accelerated genetec-security-center multiprocessing-python docker-compose-nvidia

• Updated Jun 30, 2026
• Python

SuperPoint keypoint detector optimized for edge deployment: PyTorch → ONNX → TensorRT FP16. ×ばつ speedup (14.66ms → 0.45ms, 2203 FPS) on Tesla T4. Targets Jetson Orin Ampere FP16 pipeline for real-time Visual SLAM.

computer-vision robotics cuda pytorch feature-extraction fp16 visual-slam tensorrt onnx superpoint edge-inference jetson-orin tensorrt-optimization

• Updated Jun 8, 2026
• Python

High-performance, multi-GPU computer vision pipeline for Genetec streams using NVIDIA DeepStream, Triton, and Numba JIT to track personnel, evaluate weapon threats, and perform in-VRAM OCR analytics.

object-tracking gstreamer-pipeline license-plate-recognition threat-detection ada-lovelace edge-ai triton-inference-server paddleocr pyds numba-jit nvidia-deepstream rtx-5090 nvidia-blackwell tensorrt-optimization mqtt-telemetry rtx-6000-ada cuda-accelerated genetec-security-center multiprocessing-python docker-compose-nvidia

• Updated Jul 14, 2026
• Python