YOLO11 with CUDA and TensorRT

This project provides a high-performance implementation of YOLOv11 object detection using TensorRT for inference acceleration. The pipeline processes images and videos in batches, leveraging CUDA for preprocessing and inference. Non-Maximum Suppression (NMS) and postprocessing are performed on the CPU to optimize results.

GitHub Repository

Feature

• CUDA-accelerated preprocessing and inference for faster performance.
• Batch processing for images and videos, supporting multiple inputs simultaneously.
• Postprocessing with NMS to refine detections.
• Threaded Execution to handle multiple input streams concurrently, with each stream processed on separate CUDA streams for maximum GPU utilization.
• Scalable pipeline: Process multiple files (images or videos) in parallel.
• Outputs detections with bounding boxes, class labels, and confidence scores.

Report