NVIDIA Releases DALI Library & nvJPEG GPU-Accelerated Library For JPEG Decode

For coinciding with the start of the Computer Vision and Patern Recognition conference starting this week in Utah, NVIDIA has a slew of new software announcements.

First up NVIDIA has announced the open-source DALI library for GPU-accelerated data augmentation and image loading that is optimized for data pipelines of deep learning frameworks like ResNET-50, TensorFlow, and PyTorch.

Also out but being closed-source is nvJPEG, a high-performance GPU-accelerated library for JPEG decoding. The nvJPEG library can be used for decoding of single/multiple JPEG images, various conversion updates, and more both on the CPU and GPU.

Details on DALI and nvJPEG via developer.nvidia.com.

Also out today is the DeepStream SDK 2.0 and Apex as a open-source PyTorch extension to help deep learning training performance with NVIDIA Volta GPUs.