Release Notes

VPI-4.0.0

VPI-4.0.0 is the first production release of VPI-4.0 branch. This VPI version supports Jetson AGX Thor DevKit and Jetson AGX Thor T5000.

For previous release notes refer to Previous Releases

Release Highlights

• Added vpiSubmitCopyImage, vpiSubmitCopyArray, and vpiSubmitCopyPyramid functions for copying containers
• Added VPI_ERROR_NOT_SUPPORTED status code for operations not supported on current platform
• Added VPIByte type for byte operations (closest to C++'s std::byte)
• Redesigned VPIImagePlanePitchLinear structure for better memory management
• PVA Backend: Added support for CropScaler algorithm and ORB feature detector
• VIC Backend: Added scaling limits on Jetson Thor platform
• Next Gen GPU support (Ada Lovelace (sm_89), Hopper (sm_90), Blackwell (sm_100, sm_120))
• Compatibility with CUDA13
• Compatability with PVA GEN3
• CUDA algorithms: average speedup of 1.9 over Orin AGX
• CPU algorithms: average speedup of 1.7 over Orin AGX
• VIC algorithms: average speedup of 1.9 over Orin AGX
• OFA algorithms: up to 8x faster.
• PVA: cropscaler 5 times faster. average speedup of 1.4 over Orin AGX

Other Updates

• Removed NVENC backend from the supported backends
• Removed legacy compatibility layer for vpiSubmitORBDescriptorExtractor
• Added new swizzle type VPI_DETAIL_DEF_SWIZZLE_ENUM(0, 0, X, 1)
• Redesigned VPIImagePlanePitchLinear structure:
  • Replaced void *data with VPIByte *pBase and int64_t offsetBytes
  • This allows multiple planes to share the same allocation with different offsets
• Changed bit position of VPI_ENABLE_CROSS_CHECK from bit 0 to bit 4
• Removed VPI_ACCUMULATE_OUTPUT flag (bit 3)
• PVA Backend added support for:
  • CropScaler algorithm
  • ORB feature detector
  • Various operations with specific constraints
• VIC Backend: Added scaling limits on Jetson Thor platform
• VPI_INTERP_NEAREST is not supported on Jetson Thor

Known Issues

• StereoDisparity: The following block linear formats are not working: VPI_IMAGE_FORMAT_NV12_ER_BL, VPI_IMAGE_FORMAT_NV12_ER_BL16, VPI_IMAGE_FORMAT_NV12_BL, VPI_IMAGE_FORMAT_NV12_BL16. Workaround is to convert those image formats.
• Dense Optical Flow:
  • Pyramid levels > 3 is not working.
  • There is no discernible difference in quality from low, medium to high.
• Background Subtractor: Initial frames may produce unstable background model.
• sample 11: Fisheye: please install "sudo python3 -m pip install opencv-python==4.8.0.74; sudo python3 -m pip install pillow numpy==1.26;"
• PyTorch: 16-vpi_pytorch sample fails due to the absence of a compatible PyTorch module for CUDA Toolkit 13.0.
• Sample 18: ORB: For C++, pva backend does not work. Python example does not work.
• sample 19: DCF_tracker: cuda backend works only with "nvpmodel -m 0"  setting.

Notices

Disclaimer

ALL NVIDIA DESIGN SPECIFICATIONS, REFERENCE BOARDS, FILES, DRAWINGS, DIAGNOSTICS, LISTS, AND OTHER DOCUMENTS (TOGETHER AND SEPARATELY, "MATERIALS") ARE BEING PROVIDED "AS IS." NVIDIA MAKES NO WARRANTIES, EXPRESSED, IMPLIED, STATUTORY, OR OTHERWISE WITH RESPECT TO THE MATERIALS, AND EXPRESSLY DISCLAIMS ALL IMPLIED WARRANTIES OF NONINFRINGEMENT, MERCHANTABILITY, AND FITNESS FOR A PARTICULAR PURPOSE.

Information furnished is believed to be accurate and reliable. However, NVIDIA Corporation assumes no responsibility for the consequences of use of such information or for any infringement of patents or other rights of third parties that may result from its use. No license is granted by implication of otherwise under any patent rights of NVIDIA Corporation. Specifications mentioned in this publication are subject to change without notice. This publication supersedes and replaces all other information previously supplied. NVIDIA Corporation products are not authorized as critical components in life support devices or systems without express written approval of NVIDIA Corporation.

Trademarks

NVIDIA and the NVIDIA logo are trademarks or registered trademarks of NVIDIA Corporation in the U.S. and other countries. Other company and product names may be trademarks of the respective companies with which they are associated.

Copyright

© 2019-2025 NVIDIA Corporation. All rights reserved.