JetPack SDK 4.4 DP Archive

OS

NVIDIA L4T provides the bootloader, Linux kernel, necessary firmwares, NVIDIA drivers, sample filesystem, and more.

JetPack 4.4 Developer Preview Highlights:

• Support for upgrading JetPack and L4T using debian package management tool
• Support for Generic Timestamping Engine (GTE) for Jetson AGX Xavier and Jetson Xavier NX
• Support for Dynamic Frequency Scaling (DFS) for Video Image Compositor (VIC) using actmon
• SE (Security Engine) samples to demonstrate hardware backed authentication and encryption capabilities of Jetson TX2 series, Jetson AGX Xavier and Jetson Xavier NX modules
• Utility to fuse multiple Jetson modules simultaneously
• Option to specify APP partition size on the microSD card during initial configuration at first boot of Jetson Xavier NX and Jetson Nano Developer Kits
• Supports Vulkan 1.2
• Supports DeepStream 4.0.2

TensorRT

TensorRT is a high performance deep learning inference runtime for image classification, segmentation, and object detection neural networks. TensorRT is built on CUDA, NVIDIA’s parallel programming model, and enables you to optimize inference for all deep learning frameworks. It includes a deep learning inference optimizer and runtime that delivers low latency and high-throughput for deep learning inference applications.

JetPack 4.4 Developer Preview Highlights:

• New layers and operators
• Support for mixed precision inference
• Support for loop constructs
• Support for Quantized models
• Support for Boolean tensors
• Support for Empty tensors
• Support for INT8 calibration with dynamic shapes
• ONNX parser with dynamic shapes support

cuDNN

CUDA Deep Neural Network library provides high-performance primitives for deep learning frameworks. It provides highly tuned implementations for standard routines such as forward and backward convolution, pooling, normalization, and activation layers.

JetPack 4.4 Developer Preview Highlights:

• cuDNN is now split into multiple training and inference libraries

CUDA

CUDA Toolkit provides a comprehensive development environment for C and C++ developers building GPU-accelerated applications. The toolkit includes a compiler for NVIDIA GPUs, math libraries, and tools for debugging and optimizing the performance of your applications.

JetPack 4.4 Developer Preview Highlights:

• 50% launch latency reduction for CUDA kernels resulting in improved GPU utilization and lower CPU utilization.

Multimedia API

The Jetson Multimedia API package provides low level APIs for flexible application development.

Camera application API: libargus offers a low-level frame-synchronous API for camera applications, with per frame camera parameter control, multiple (including synchronized) camera support, and EGL stream outputs. RAW output CSI cameras needing ISP can be used with either libargus or GStreamer plugin. In either case, the V4L2 media-controller sensor driver API is used.

Sensor driver API: V4L2 API enables video decode, encode, format conversion and scaling functionality. V4L2 for encode opens up many features like bit rate control, quality presets, low latency encode, temporal tradeoff, motion vector maps, and more.

Computer Vision

VisionWorks is a software development package for Computer Vision (CV) and image processing.

OpenCV is a leading open source library for computer vision, image processing and machine learning.

VPI (Vision Programing Interface)1, a software library that provides Computer Vision / Image Processing algorithms implemented on PVA2 (Programmable Vision Accelerator), GPU and CPU

JetPack 4.4 Developer Preview Highlights:

• Performance optimization of algorithms introduced in VPI 0.1.0: up to 45x on GPU and up to 90x on CPU backends
• New Image FFT, Image iFFT and Image Format converter algorithms added with support for CPU and GPU backends. (PVA backend will be supported in a future release.)

Developer Tools

CUDA Toolkit provides a comprehensive development environment for C and C++ developers building high-performance GPU-accelerated applications with CUDA libraries. The toolkit includes Nsight Eclipse Edition, debugging and profiling tools including Nsight Compute, and a toolchain for cross-compiling applications.

NVIDIA Nsight Systems is a low overhead system-wide profiling tool, providing the insights developers need to analyze and optimize software performance.

NVIDIA Nsight Graphics is a standalone application for debugging and profiling graphics applications.

JetPack 4.4 Developer Preview Highlights:

• NVIDIA Nsight Systems
  • Enhanced data analysis with option to export to SQLite, HDF5 or JSON
  • Support for sampling Xavier PMU extensions
  • Reduced NVTX overhead
  • New CLI support for profiling on devices with intermittent network connectivity
• NVIDIA Nsight Graphics
  • Support to save and load custom named layouts
  • Improved events view display and filtering
  • Enhanced support for mixed DPI monitor scaling
  • Support for new Vulkan Extensions

DeepStream SDK

NVIDIA’s DeepStream SDK delivers a complete streaming analytics toolkit for AI-based multi-sensor processing, video and image understanding. DeepStream is an integral part of NVIDIA Metropolis, the platform for building end-to-end services and solutions that transform pixel and sensor data to actionable insights.

DeepStream SDK 5.0 developer preview highlights

• Run popular Deep Learning framework natively with DeepStream: New inference capability with Triton Inference Server (previously TensorRT Inference Server) enables developers to deploy a model natively in TensorFlow, TensorFlow-TensorRT, PyTorch, or ONNX in the DeepStream pipeline using NVIDIA Triton Inference server.
• Python based development: Select from either C/C++ or Python based applications for your use-case. We’ve introduced Python support with seven reference apps . Both development options provide comparable performance.
• Secure communication between edge and cloud using Kafka message broker and SSL communication
• Introducing Jetson Xavier NX support: Deploy AI applications on the world’s smallest AI supercomputer at the edge. Reference apps are common among all platforms.
• IoT capabilities
• Superior DeepStream app control from edge or cloud with bi-directional IoT messaging
• Dynamic AI model update on the go reduces app downtime
• Interoperability with TAO Toolkit