nvJPEG is a high-performance GPU-accelerated library designed for decoding, encoding, and transcoding JPEG format images. It provides low-latency processing for common JPEG formats used in computer vision applications such as image classification, object detection, and image segmentation. The library leverages both CPU and GPU for hybrid decoding, ensuring high throughput and low latency.

Features

• Hybrid Decoding: Utilizes both CPU and GPU for efficient JPEG decoding.
• Hardware Acceleration: Supports baseline JPEG decode on A100 GPUs.
• Flexible Decoding Options: Offers single image and batched image decoding, as well as single phase and multiphase decoding.
• Color Space Conversion: Converts images to RGB, BGR, RGBI, BGRI, and YUV color spaces.
• Memory Management: Input is in host memory, and output is in GPU memory, with user-provided memory manager for device and pinned host memory allocations.

Benefits

• High Throughput: Delivers higher throughput and lower latency compared to CPU-only decoding.
• Versatility: Suitable for various computer vision applications, including image classification, object detection, and image segmentation.
• Efficiency: Reduces processing time and improves performance by leveraging GPU acceleration.