NVIDIA DeepStream SDK 2.0 for Tesla now available

Accelerated Computing Intelligent Video Analytics DeepStream SDK
1

Today, we announced the availability of DeepStream SDK 2.0 for Tesla. The NVIDIA DeepStream SDK is ideal for developers who are creating and deploying AI-based solutions for video analytics applications at scale, offering a complete framework and all the essential building blocks. It lets you focus on the core deep learning networks and IP you care about, rather than designing an end-to-end solution from scratch.

DeepStream makes use of:

  • NVIDIA® TensorRT™ and NVIDIA CUDA® for AI and other GPU computing tasks
  • Video CODEC SDK and multimedia APIs for accelerated encode and decode
  • Imaging APIs for capture and processing
  • A graph-based architecture and modular plug-ins to create configurable processing pipelines

Download DeepStream SDK for Tesla today

2

Thank you for this distribution.

I went through the download/installation process, but I am having an issue running the example application as follows:

** ERROR: <create_multi_source_bin:347>: Failed to create element ‘src_bin_muxer’
** ERROR: <create_multi_source_bin:402>: create_multi_source_bin failed
** ERROR: <create_pipeline:949>: create_pipeline failed
** ERROR: main:469: Failed to create pipeline
Quitting

(deepstream-app:2833): GStreamer-CRITICAL **: gst_element_get_static_pad: assertion ‘GST_IS_ELEMENT (element)’ failed

(deepstream-app:2833): GStreamer-CRITICAL **: gst_pad_send_event: assertion ‘GST_IS_PAD (pad)’ failed
App run failed

Do you have any pointers? I already tried removing the gstreamer cache as mentioned in the README.

Thank You.

3

Hi,

The picture(NVIDIA_DeepStream_SDK_on_Tesla_V2.0_User_Guide.pdf, page 4) is very vague and I can’t see the text above.

Can you provide a clear version of this picture?

Thank You.

4

Must the model be caffe model in the DeepStream SDK 2.0 for Tesla?
How to run the tensorflow model in the DeepStream SDK 2.0 for Tesla?

5

Hi,

thank you for this release.

I agree with @suchi, the picture is unreadable. Please change it.

I have additional question about the status of 1.5 version, where you are using FFMPEG as a backend. Is FFMPEG version of DeepStream deprecated? Will there be any updates on FFMPEG version of DeepStream. I find FFMPEG version (i.e. 1.5) much more capable in terms of the flexibility as is easier to use (developer wise, as you work directly in C++ with relatively simple API) compared to 2.0 (GStreamer) where you are very limited with the capabilities that you can add and especially how easily you can add it. There is very little source code in 2.0, you are more or less limited to txt configuration files. Can you provide source code for the pre-built sample application binaries in order to get more insight? (Similar as source code is provided for 1.5 FFMPEG).

6

Hi,

We have downloaded the DeepStream 2.0, and study with it now.

Thanks.

7

Guys,

Thanks for the feedback. We will fix the unreadable text inside page 4 of NVIDIA_DeepStream_SDK_on_Tesla_V2.0_User_Guide.pdf.

Regarding specific issue while using DeepStream SDK, please file a topic separately.

Thanks

8

Hi dejan.stepec,

Regarding FFMPEG usage:

  1. For DS 2.0 we have stopped supporting FFMPEG and migrated to Gstreamer for scalability, flexibility and plug and play arch.
  2. The query related to providing source for reference application, yes we intend to provide the source for reference application once code clean is completed. For now any user can use “deepstream-test1 and deepstream-test2” sources for understanding and implementation.

Thanks

9

Is there a time frame for 2.0 on Jetson TX2?

10

I was running the deepstream2.o samples and it is not getting compiled.
Thinking that it would be a gstreamer issue I reinstalled it, but it didnt helped.
Here is the error that I am getting for
deepstream-app -c source4_720p_dec_infer-resnet_tracker_sgie_tiled_display_int8.txt

Using Cached GIE model /home/graymatics/Deep-stream/DeepStream_Release/samples/configs/deepstream-app/…/…/models/Secondary_CarMake/resnet18.caffemodel_b16_int8.cache crypto flags(0)
Using Cached GIE model /home/graymatics/Deep-stream/DeepStream_Release/samples/configs/deepstream-app/…/…/models/Secondary_CarColor/resnet18.caffemodel_b16_int8.cache crypto flags(0)
Using Cached GIE model /home/graymatics/Deep-stream/DeepStream_Release/samples/configs/deepstream-app/…/…/models/Secondary_VehicleTypes/resnet18.caffemodel_b16_int8.cache crypto flags(0)
Using Cached GIE model /home/graymatics/Deep-stream/DeepStream_Release/samples/configs/deepstream-app/…/…/models/Primary_Detector/resnet10.caffemodel_b4_int8.cache crypto flags(0)

Runtime commands:
h: Print this help
q: Quit

p: Pause
r: Resume

NOTE: To expand a source in the 2D tiled display and view object details, left-click on the source.
To go back to the tiled display, right-click anywhere on the window.

** INFO: <bus_callback:75>: Pipeline ready

nvstreamtiler: batchSize set as 4
cuGraphicsGLRegisterBuffer failed with error(304) gst_eglglessink_cuda_init texture = 1
** INFO: <bus_callback:60>: Pipeline running

**PERF: FPS 0 (Avg) FPS 1 (Avg) FPS 2 (Avg) FPS 3 (Avg)
**PERF: 144.80 (144.80) 144.80 (144.80) 144.80 (144.80) 144.80 (144.80)
**PERF: -nan (144.80) -nan (144.80) -nan (144.80) -nan (144.80)
Error from secondary_gie_bin_queue: Internal data flow error.
Debug info: gstqueue.c(992): gst_queue_handle_sink_event (): /GstPipeline:pipeline/GstBin:secondary_gie_bin/GstQueue:secondary_gie_bin_queue:
streaming task paused, reason not-negotiated (-4)
Quitting
App run failed

Same with all of the configs>

Is there anything wrong with the setup?
I am using P4 Tesla and installed sudo apt install libssl1.0.0 libjpeg8 libgstreamer1.0-0 gstreamer-tools gstreamer1.0-plugins-good gstreamer1.0-plugins-bad gstreamer1.0-plugins-ugly gstreamer1.0-libav libgtk-3-0. But no hope; opencv is compiled with ffmpeg which is further compiled with binary h264 and nv code. Let me know if you want to know further about the setup.