I’m trying to use the ZED SDK on my machine with my ZED 2. Here is my current setup that has been installed successfully:
OS: Ubuntu 24.04
CUDA version 12.8
cuDNN version 9.8
TensorRT version 10.9
ZED SDK 5.0
The installation of the ZED SDK has been completed without installing the SDK Python API since on Ubuntu 24.04 gives problems (you can’t install pip packages system-wide, it throws errors if you try to do it).
Running ZED Diagnostics everything seems fine.
Here are the errors when I try to run ZEDfu, ZED_Depth_Viewer and the spatial mapping sample.
ZEDfu
ubm@ubm:~/Downloads$ ZEDfu
Stack trace (most recent call last):
#6 Object "[0xffffffffffffffff]", at 0xffffffffffffffff, in
#5 Object "ZEDfu", at 0x5a6631408a64, in
#4 Object "/lib/x86_64-linux-gnu/libc.so.6", at 0x7064ba82a28a, in __libc_start_main
#3 Object "/lib/x86_64-linux-gnu/libc.so.6", at 0x7064ba82a1c9, in
#2 Object "ZEDfu", at 0x5a6631408954, in
#1 Object "ZEDfu", at 0x5a663142e21e, in
#0 Object "/lib/x86_64-linux-gnu/libQt5Gui.so.5", at 0x7064bbb63f23, in QStandardItem::flags() const
Segmentation fault (Address not mapped to object [(nil)])
Segmentation fault (core dumped)
ZED_Depth_Viewer
ubm@ubm:~/Downloads$ ZED_Depth_Viewer
Downloading: NEURAL DEPTH...
[2025-04-17 17:39:52 UTC][ZED][INFO] AI model not found, downloading
/usr/local/zed/resources/n 100%[========================================>] 26.36M 394KB/s in 45s
Optimizing: NEURAL DEPTH...
[2025-04-17 17:40:39 UTC][ZED][INFO] Please wait while the AI model is being optimized for your graphics card
This operation will be run only once and may take a few minutes
Optimizing neural_depth_5.2 | 100.0%[=============================>] Done ! in 2min 9s
[2025-04-17 17:42:55 UTC][ZED][INFO] Logging level INFO
[2025-04-17 17:42:55 UTC][ZED][INFO] [Init] Depth mode: NEURAL
[2025-04-17 17:42:56 UTC][ZED][INFO] [Init] Camera successfully opened.
[2025-04-17 17:42:56 UTC][ZED][INFO] [Init] Camera FW version: 1523
[2025-04-17 17:42:56 UTC][ZED][INFO] [Init] Video mode: HD720@30
[2025-04-17 17:42:56 UTC][ZED][INFO] [Init] Serial Number: S/N 29901994
[2025-04-17 17:42:56 UTC][ZED][INFO] [Init] No calibration file found for SN 29901994. Downloading...
[2025-04-17 17:42:57 UTC][ZED][INFO] [Init] Calibration file downloaded.
in bool ImageHandler::initialize(sl::Mat&) : Err [999]: unknown error.
in bool ImageHandler::initialize(sl::Mat&) : Err [400]: invalid resource handle.
in bool ImageHandler::initialize(sl::Mat&) : Err [400]: invalid resource handle.
in bool ImageHandler::initialize(sl::Mat&) : Err [400]: invalid resource handle.
in bool ImageHandler::initialize(sl::Mat&) : Err [999]: unknown error.
in bool ImageHandler::initialize(sl::Mat&) : Err [400]: invalid resource handle.
in bool ImageHandler::initialize(sl::Mat&) : Err [400]: invalid resource handle.
in bool ImageHandler::initialize(sl::Mat&) : Err [400]: invalid resource handle.
in void OGLWidget::Deinit() : Err [400]: invalid resource handle.
in virtual void InputThreadModul::run() : Err [400]: invalid resource handle.
Spatial mapping sample
ubm@ubm:/usr/local/zed/samples/spatial mapping/spatial mapping/cpp/build$ ./ZED_Spatial_Mapping
[2025-04-17 17:44:41 UTC][ZED][INFO] Logging level INFO
[2025-04-17 17:44:42 UTC][ZED][INFO] Using USB input... Switched to default resolution HD720
[2025-04-17 17:44:42 UTC][ZED][INFO] [Init] Depth mode: NEURAL
[2025-04-17 17:44:42 UTC][ZED][INFO] [Init] Camera successfully opened.
[2025-04-17 17:44:42 UTC][ZED][INFO] [Init] Camera FW version: 1523
[2025-04-17 17:44:42 UTC][ZED][INFO] [Init] Video mode: HD720@30
[2025-04-17 17:44:42 UTC][ZED][INFO] [Init] Serial Number: S/N 29901994
Shortcuts
- 'l' to enable/disable current live point cloud display
- 'w' to switch mesh display from faces to triangles
- 'd' to switch background color from dark to light
Error: CUDA - OpenGL Interop failed (999)
Segmentation fault (core dumped)
Basically, I’m on a clean installation of Ubuntu on my desktop PC with an RTX 4060 Ti.
When executing glxinfo | grep OpenGL i get indeed OpenGL vendor string: Intel (since I’m using an Intel CPU with integrated graphics). If I run __NV_PRIME_RENDER_OFFLOAD=1 __GLX_VENDOR_LIBRARY_NAME=nvidia ZED_Depth_Viewer, then the neural model is loaded successfully and the GUI as well, and everything works.
If I run __NV_PRIME_RENDER_OFFLOAD=1 __GLX_VENDOR_LIBRARY_NAME=nvidia ZEDfu, the error remains the same.
Instead, if I install CUDA 12.6 and ZED SDK 4.2.5, ZEDfu manages to start up. And again, with SDK 4.2.5, __NV_PRIME_RENDER_OFFLOAD=1 __GLX_VENDOR_LIBRARY_NAME=nvidia ZEDfu works and I can even use the neural models
It seems that ZEDfu for SDK 5.0.0 has been built with Qt5 or other shared libraries not compatible with the standard Ubuntu 24 ones (or I messed up the installation of Qt or other libraries, even though I only followed the recommended steps), independently of NVIDIA driver usage (offload) or not. Another proof of that is that ZEDfu 5.0.0 with CUDA 12.6 throws the same errors.
I verified that the Qt5 libs are installed in the system folders. For completeness, I provide the ldd output of ZEDfu 5.0.0: