NVIDIA DeepStream SDK
THIS IS AN OLD VERSION OF DEEPSTREAM SDK, WHICH USES DIFFERENT ARCHITECTURE FROM LATER VERSIONS. THIS VERSION IS EASILY TO USE BUT ONLY USES TENSORRT 3 AND DOES NOT SUPPORT MANY LASTEST DEEP LEARNING MODELS.
DeepStream provides an easy-to-use and high-performance SDK for video content analysis, which simplifies development of high-performance video analytics applications powered by deep learning. DeepStream enables customer to make optimum use of underlying GPU architectures, including hardware decoding support, thereby achieving high levels of efficiency, performance, and scale. Furthermore, DeepStream provides a flexible plug-in mechanism for the user to incorporate their custom functionality to video analytics applications to meet their unique needs.
DeepStream provides a high-level C++ API for GPU-accelerated video decoding, inference. DeepStream is built on the top of NVCODEC and NVIDIA® TensorRT™, which are responsible for video decoding and deep learning inference, respectively.
The following are the key features of DeepStream:
DeepStream has the following software dependencies:
NVIDIA recommends that DeepStream be run on a hardware platform with an NVIDIA Tesla® P4 or P40 graphics card.
The DeepStream SDK consists of two main parts: the library and the workflow demonstration samples. The installed DeepStream package includes the directories /lib, /include, /doc, and /samples.
In Progress