Skip to main content

[FLaNK] Streaming EdgeAI on the new NVIDIA Jetson Nano 2GB with MiNiFi Agents To FLaNK Applications

 [FLaNK] Streaming EdgeAI on the new NVIDIA Jetson Nano 2GB with MiNiFi Agents To FLaNK Applications

Plug Into Community AI Apps:  https://youtu.be/2T8CG7lDkcU

I am not patient enough to shoot an unboxing video, I was too excited to get this superb machine running.   The NVIDIA Jetson Nano 2GB is now available for purchase for only $59!!!


The 2GB version of NVIDIA Jetson Nano is great, you really don't miss anything that was removed.   I have copied over my MiNiFi agent and code from other Jetson Nanos, Xavier NX and TX1 and it all works fine.   The speed is fine for most needs especially for development and prototyping.   I prefer the Xavier, but at this price you can't go wrong.   I am definitely going to be getting Jetson Nanos instead of other devices for most IoT / Edge AI use cases.   I have used my NVidia Jetson 2GB for demos for a number of events including ApacheCon, BeamSummit, Open Source Summit and AI Dev World.


I installed the fswebcam to capture still images and build up a directory of them to process.


You must install and run:   https://github.com/dusty-nv/jetson-inference.   You get great libraries, tutorials, documentation and examples.   I usually build my apps starting from one of these examples and use one of the excellent NVIDIA pre-built models.   This rapidly accelerates my development and deployment of EdgeAI applications whether they are IoT or other purposes.   This is working with standard Raspberry Pi plug in cameras and the excellent Logitech USB web cameras that I have used with all my other NVIDIA devices.

At this price point, there seems no reason that every developer in every company should have one.   It's a great place to test out Edge AI applications and run classifications at a decent speed.   This is a real machine despite it.

I was facilitating data journeys at the NetHope Global Summit today and I thought these $59 devices could be great for non-profits to use for many data collection and analytics purposes in the field.  https://www.nethopeglobalsummit.org/agenda-2020#sz-tab-44134 I am exploring some use cases to see if I can pre build some easy applications that an NGO could just pick up and run with.   Let's see what develops.   A $59 GPU edge device enables some new applications at an affordable cost.   $59 won't give me a lot of cloud, but I can get a powerful small data collection device that runs ML, DL, cameras, MiNiFi Agents, Python and Java.   With 2 Gigabytes of fast RAM and a GPU, one is limited by their imagination.


Example Application



Output Exampl Datae:

{"uuid": "nano_uuid_cmq_20201026202757", "ipaddress": "192.168.1.169", "networktime": 47.7275505065918, "detectleft": 1.96746826171875, "detectconfidence": 52.86550521850586, "cputemp": "34.0", "gputemp": "30.0", "gputempf": "86", "cputempf": "93", "runtime": "169", "host": "nano5", "filename": "/opt/demo/images/out_iue_20201026202757.jpg", "host_name": "nano5", "macaddress": "00:e0:4c:49:d8:b7", "end": "1603744246.924455", "te": "169.4200084209442", "systemtime": "10/26/2020 16:30:46", "cpu": 9.9, "diskusage": "37100.4 MB", "memory": 91.5, "id": "20201026202757_64d69a82-88d8-45f8-be06-1b836cb6cc84"}


Below is some example output for running a Python script to classify a webcamera (a low end Logi webcam, but you can use a Raspberry Pi camera).   We would be best served by running this continuously outputting log messages and images for MiNiFi agents to scoop up and send to a server for routing, transformation and processing.


root@nano5:/opt/demo/minifi-jetson-nano# jetson_clocks 
root@nano5:/opt/demo/minifi-jetson-nano# python3 detect.py 
[gstreamer] initialized gstreamer, version 1.14.5.0
[gstreamer] gstCamera -- attempting to create device v4l2:///dev/video0
[gstreamer] gstCamera -- found v4l2 device: HD Webcam C615
[gstreamer] v4l2-proplist, device.path=(string)/dev/video0, udev-probed=(boolean)false, device.api=(string)v4l2, v4l2.device.driver=(string)uvcvideo, v4l2.device.card=(string)"HD\ Webcam\ C615", v4l2.device.bus_info=(string)usb-70090000.xusb-3.2, v4l2.device.version=(uint)264588, v4l2.device.capabilities=(uint)2216689665, v4l2.device.device_caps=(uint)69206017;
[gstreamer] gstCamera -- found 30 caps for v4l2 device /dev/video0
[gstreamer] [0] video/x-raw, format=(string)YUY2, width=(int)1920, height=(int)1080, pixel-aspect-ratio=(fraction)1/1, framerate=(fraction)5/1;
[gstreamer] [1] video/x-raw, format=(string)YUY2, width=(int)1600, height=(int)896, pixel-aspect-ratio=(fraction)1/1, framerate=(fraction){ 15/2, 5/1 };
[gstreamer] [2] video/x-raw, format=(string)YUY2, width=(int)1280, height=(int)720, pixel-aspect-ratio=(fraction)1/1, framerate=(fraction){ 10/1, 15/2, 5/1 };
[gstreamer] [3] video/x-raw, format=(string)YUY2, width=(int)960, height=(int)720, pixel-aspect-ratio=(fraction)1/1, framerate=(fraction){ 15/1, 10/1, 15/2, 5/1 };
[gstreamer] [4] video/x-raw, format=(string)YUY2, width=(int)1024, height=(int)576, pixel-aspect-ratio=(fraction)1/1, framerate=(fraction){ 15/1, 10/1, 15/2, 5/1 };
[gstreamer] [5] video/x-raw, format=(string)YUY2, width=(int)800, height=(int)600, pixel-aspect-ratio=(fraction)1/1, framerate=(fraction){ 24/1, 20/1, 15/1, 10/1, 15/2, 5/1 };
[gstreamer] [6] video/x-raw, format=(string)YUY2, width=(int)864, height=(int)480, pixel-aspect-ratio=(fraction)1/1, framerate=(fraction){ 24/1, 20/1, 15/1, 10/1, 15/2, 5/1 };
[gstreamer] [7] video/x-raw, format=(string)YUY2, width=(int)800, height=(int)448, pixel-aspect-ratio=(fraction)1/1, framerate=(fraction){ 30/1, 24/1, 20/1, 15/1, 10/1, 15/2, 5/1 };
[gstreamer] [8] video/x-raw, format=(string)YUY2, width=(int)640, height=(int)480, pixel-aspect-ratio=(fraction)1/1, framerate=(fraction){ 30/1, 24/1, 20/1, 15/1, 10/1, 15/2, 5/1 };
[gstreamer] [9] video/x-raw, format=(string)YUY2, width=(int)640, height=(int)360, pixel-aspect-ratio=(fraction)1/1, framerate=(fraction){ 30/1, 24/1, 20/1, 15/1, 10/1, 15/2, 5/1 };
[gstreamer] [10] video/x-raw, format=(string)YUY2, width=(int)432, height=(int)240, pixel-aspect-ratio=(fraction)1/1, framerate=(fraction){ 30/1, 24/1, 20/1, 15/1, 10/1, 15/2, 5/1 };
[gstreamer] [11] video/x-raw, format=(string)YUY2, width=(int)352, height=(int)288, pixel-aspect-ratio=(fraction)1/1, framerate=(fraction){ 30/1, 24/1, 20/1, 15/1, 10/1, 15/2, 5/1 };
[gstreamer] [12] video/x-raw, format=(string)YUY2, width=(int)320, height=(int)240, pixel-aspect-ratio=(fraction)1/1, framerate=(fraction){ 30/1, 24/1, 20/1, 15/1, 10/1, 15/2, 5/1 };
[gstreamer] [13] video/x-raw, format=(string)YUY2, width=(int)176, height=(int)144, pixel-aspect-ratio=(fraction)1/1, framerate=(fraction){ 30/1, 24/1, 20/1, 15/1, 10/1, 15/2, 5/1 };
[gstreamer] [14] video/x-raw, format=(string)YUY2, width=(int)160, height=(int)120, pixel-aspect-ratio=(fraction)1/1, framerate=(fraction){ 30/1, 24/1, 20/1, 15/1, 10/1, 15/2, 5/1 };
[gstreamer] [15] image/jpeg, width=(int)1920, height=(int)1080, pixel-aspect-ratio=(fraction)1/1, framerate=(fraction){ 30/1, 24/1, 20/1, 15/1, 10/1, 15/2, 5/1 };
[gstreamer] [16] image/jpeg, width=(int)1600, height=(int)896, pixel-aspect-ratio=(fraction)1/1, framerate=(fraction){ 30/1, 24/1, 20/1, 15/1, 10/1, 15/2, 5/1 };
[gstreamer] [17] image/jpeg, width=(int)1280, height=(int)720, pixel-aspect-ratio=(fraction)1/1, framerate=(fraction){ 30/1, 24/1, 20/1, 15/1, 10/1, 15/2, 5/1 };
[gstreamer] [18] image/jpeg, width=(int)960, height=(int)720, pixel-aspect-ratio=(fraction)1/1, framerate=(fraction){ 30/1, 24/1, 20/1, 15/1, 10/1, 15/2, 5/1 };
[gstreamer] [19] image/jpeg, width=(int)1024, height=(int)576, pixel-aspect-ratio=(fraction)1/1, framerate=(fraction){ 30/1, 24/1, 20/1, 15/1, 10/1, 15/2, 5/1 };
[gstreamer] [20] image/jpeg, width=(int)800, height=(int)600, pixel-aspect-ratio=(fraction)1/1, framerate=(fraction){ 30/1, 24/1, 20/1, 15/1, 10/1, 15/2, 5/1 };
[gstreamer] [21] image/jpeg, width=(int)864, height=(int)480, pixel-aspect-ratio=(fraction)1/1, framerate=(fraction){ 30/1, 24/1, 20/1, 15/1, 10/1, 15/2, 5/1 };
[gstreamer] [22] image/jpeg, width=(int)800, height=(int)448, pixel-aspect-ratio=(fraction)1/1, framerate=(fraction){ 30/1, 24/1, 20/1, 15/1, 10/1, 15/2, 5/1 };
[gstreamer] [23] image/jpeg, width=(int)640, height=(int)480, pixel-aspect-ratio=(fraction)1/1, framerate=(fraction){ 30/1, 24/1, 20/1, 15/1, 10/1, 15/2, 5/1 };
[gstreamer] [24] image/jpeg, width=(int)640, height=(int)360, pixel-aspect-ratio=(fraction)1/1, framerate=(fraction){ 30/1, 24/1, 20/1, 15/1, 10/1, 15/2, 5/1 };
[gstreamer] [25] image/jpeg, width=(int)432, height=(int)240, pixel-aspect-ratio=(fraction)1/1, framerate=(fraction){ 30/1, 24/1, 20/1, 15/1, 10/1, 15/2, 5/1 };
[gstreamer] [26] image/jpeg, width=(int)352, height=(int)288, pixel-aspect-ratio=(fraction)1/1, framerate=(fraction){ 30/1, 24/1, 20/1, 15/1, 10/1, 15/2, 5/1 };
[gstreamer] [27] image/jpeg, width=(int)320, height=(int)240, pixel-aspect-ratio=(fraction)1/1, framerate=(fraction){ 30/1, 24/1, 20/1, 15/1, 10/1, 15/2, 5/1 };
[gstreamer] [28] image/jpeg, width=(int)176, height=(int)144, pixel-aspect-ratio=(fraction)1/1, framerate=(fraction){ 30/1, 24/1, 20/1, 15/1, 10/1, 15/2, 5/1 };
[gstreamer] [29] image/jpeg, width=(int)160, height=(int)120, pixel-aspect-ratio=(fraction)1/1, framerate=(fraction){ 30/1, 24/1, 20/1, 15/1, 10/1, 15/2, 5/1 };
[gstreamer] gstCamera -- selected device profile:  codec=mjpeg format=unknown width=1280 height=720
[gstreamer] gstCamera pipeline string:
[gstreamer] v4l2src device=/dev/video0 ! image/jpeg, width=(int)1280, height=(int)720 ! jpegdec ! video/x-raw ! appsink name=mysink
[gstreamer] gstCamera successfully created device v4l2:///dev/video0
[gstreamer] opening gstCamera for streaming, transitioning pipeline to GST_STATE_PLAYING
[gstreamer] gstreamer changed state from NULL to READY ==> mysink
[gstreamer] gstreamer changed state from NULL to READY ==> capsfilter1
[gstreamer] gstreamer changed state from NULL to READY ==> jpegdec0
[gstreamer] gstreamer changed state from NULL to READY ==> capsfilter0
[gstreamer] gstreamer changed state from NULL to READY ==> v4l2src0
[gstreamer] gstreamer changed state from NULL to READY ==> pipeline0
[gstreamer] gstreamer changed state from READY to PAUSED ==> capsfilter1
[gstreamer] gstreamer changed state from READY to PAUSED ==> jpegdec0
[gstreamer] gstreamer changed state from READY to PAUSED ==> capsfilter0
[gstreamer] gstreamer stream status CREATE ==> src
[gstreamer] gstreamer changed state from READY to PAUSED ==> v4l2src0
[gstreamer] gstreamer changed state from READY to PAUSED ==> pipeline0
[gstreamer] gstreamer stream status ENTER ==> src
[gstreamer] gstreamer message new-clock ==> pipeline0
[gstreamer] gstreamer message stream-start ==> pipeline0
[gstreamer] gstreamer changed state from PAUSED to PLAYING ==> capsfilter1
[gstreamer] gstreamer changed state from PAUSED to PLAYING ==> jpegdec0
[gstreamer] gstreamer changed state from PAUSED to PLAYING ==> capsfilter0
[gstreamer] gstreamer changed state from PAUSED to PLAYING ==> v4l2src0
[gstreamer] gstCamera -- onPreroll
[gstreamer] gstCamera -- map buffer size was less than max size (1382400 vs 1382407)
[gstreamer] gstCamera recieve caps:  video/x-raw, format=(string)I420, width=(int)1280, height=(int)720, interlace-mode=(string)progressive, multiview-mode=(string)mono, multiview-flags=(GstVideoMultiviewFlagsSet)0:ffffffff:/right-view-first/left-flipped/left-flopped/right-flipped/right-flopped/half-aspect/mixed-mono, pixel-aspect-ratio=(fraction)1/1, chroma-site=(string)mpeg2, colorimetry=(string)1:4:0:0, framerate=(fraction)30/1
[gstreamer] gstCamera -- recieved first frame, codec=mjpeg format=i420 width=1280 height=720 size=1382407
RingBuffer -- allocated 4 buffers (1382407 bytes each, 5529628 bytes total)
[gstreamer] gstreamer changed state from READY to PAUSED ==> mysink
[gstreamer] gstreamer message async-done ==> pipeline0
[gstreamer] gstreamer changed state from PAUSED to PLAYING ==> mysink
[gstreamer] gstreamer changed state from PAUSED to PLAYING ==> pipeline0
RingBuffer -- allocated 4 buffers (14745600 bytes each, 58982400 bytes total)
jetson.inference -- detectNet loading build-in network 'ssd-mobilenet-v2'

detectNet -- loading detection network model from:
          -- model        networks/SSD-Mobilenet-v2/ssd_mobilenet_v2_coco.uff
          -- input_blob   'Input'
          -- output_blob  'NMS'
          -- output_count 'NMS_1'
          -- class_labels networks/SSD-Mobilenet-v2/ssd_coco_labels.txt
          -- threshold    0.500000
          -- batch_size   1

[TRT]    TensorRT version 7.1.3
[TRT]    loading NVIDIA plugins...
[TRT]    Registered plugin creator - ::GridAnchor_TRT version 1
[TRT]    Registered plugin creator - ::NMS_TRT version 1
[TRT]    Registered plugin creator - ::Reorg_TRT version 1
[TRT]    Registered plugin creator - ::Region_TRT version 1
[TRT]    Registered plugin creator - ::Clip_TRT version 1
[TRT]    Registered plugin creator - ::LReLU_TRT version 1
[TRT]    Registered plugin creator - ::PriorBox_TRT version 1
[TRT]    Registered plugin creator - ::Normalize_TRT version 1
[TRT]    Registered plugin creator - ::RPROI_TRT version 1
[TRT]    Registered plugin creator - ::BatchedNMS_TRT version 1
[TRT]    Could not register plugin creator -  ::FlattenConcat_TRT version 1
[TRT]    Registered plugin creator - ::CropAndResize version 1
[TRT]    Registered plugin creator - ::DetectionLayer_TRT version 1
[TRT]    Registered plugin creator - ::Proposal version 1
[TRT]    Registered plugin creator - ::ProposalLayer_TRT version 1
[TRT]    Registered plugin creator - ::PyramidROIAlign_TRT version 1
[TRT]    Registered plugin creator - ::ResizeNearest_TRT version 1
[TRT]    Registered plugin creator - ::Split version 1
[TRT]    Registered plugin creator - ::SpecialSlice_TRT version 1
[TRT]    Registered plugin creator - ::InstanceNormalization_TRT version 1
[TRT]    detected model format - UFF  (extension '.uff')
[TRT]    desired precision specified for GPU: FASTEST
[TRT]    requested fasted precision for device GPU without providing valid calibrator, disabling INT8
[TRT]    native precisions detected for GPU:  FP32, FP16
[TRT]    selecting fastest native precision for GPU:  FP16
[TRT]    attempting to open engine cache file /usr/local/bin/networks/SSD-Mobilenet-v2/ssd_mobilenet_v2_coco.uff.1.1.7103.GPU.FP16.engine
[TRT]    loading network plan from engine cache... /usr/local/bin/networks/SSD-Mobilenet-v2/ssd_mobilenet_v2_coco.uff.1.1.7103.GPU.FP16.engine
[TRT]    device GPU, loaded /usr/local/bin/networks/SSD-Mobilenet-v2/ssd_mobilenet_v2_coco.uff
[TRT]    Deserialize required 2384046 microseconds.
[TRT]    
[TRT]    CUDA engine context initialized on device GPU:
[TRT]       -- layers       117
[TRT]       -- maxBatchSize 1
[TRT]       -- workspace    0
[TRT]       -- deviceMemory 35449344
[TRT]       -- bindings     3
[TRT]       binding 0
                -- index   0
                -- name    'Input'
                -- type    FP32
                -- in/out  INPUT
                -- # dims  3
                -- dim #0  3 (SPATIAL)
                -- dim #1  300 (SPATIAL)
                -- dim #2  300 (SPATIAL)
[TRT]       binding 1
                -- index   1
                -- name    'NMS'
                -- type    FP32
                -- in/out  OUTPUT
                -- # dims  3
                -- dim #0  1 (SPATIAL)
                -- dim #1  100 (SPATIAL)
                -- dim #2  7 (SPATIAL)
[TRT]       binding 2
                -- index   2
                -- name    'NMS_1'
                -- type    FP32
                -- in/out  OUTPUT
                -- # dims  3
                -- dim #0  1 (SPATIAL)
                -- dim #1  1 (SPATIAL)
                -- dim #2  1 (SPATIAL)
[TRT]    
[TRT]    binding to input 0 Input  binding index:  0
[TRT]    binding to input 0 Input  dims (b=1 c=3 h=300 w=300) size=1080000
[TRT]    binding to output 0 NMS  binding index:  1
[TRT]    binding to output 0 NMS  dims (b=1 c=1 h=100 w=7) size=2800
[TRT]    binding to output 1 NMS_1  binding index:  2
[TRT]    binding to output 1 NMS_1  dims (b=1 c=1 h=1 w=1) size=4
[TRT]    
[TRT]    device GPU, /usr/local/bin/networks/SSD-Mobilenet-v2/ssd_mobilenet_v2_coco.uff initialized.
[TRT]    W = 7  H = 100  C = 1
[TRT]    detectNet -- maximum bounding boxes:  100
[TRT]    detectNet -- loaded 91 class info entries
[TRT]    detectNet -- number of object classes:  91
detected 0 objects in image

[TRT]    ------------------------------------------------
[TRT]    Timing Report /usr/local/bin/networks/SSD-Mobilenet-v2/ssd_mobilenet_v2_coco.uff
[TRT]    ------------------------------------------------
[TRT]    Pre-Process   CPU   0.07802ms  CUDA   0.48875ms
[TRT]    Network       CPU  45.52254ms  CUDA  44.93750ms
[TRT]    Post-Process  CPU   0.03193ms  CUDA   0.03177ms
[TRT]    Total         CPU  45.63248ms  CUDA  45.45802ms
[TRT]    ------------------------------------------------

[TRT]    note -- when processing a single image, run 'sudo jetson_clocks' before
                to disable DVFS for more accurate profiling/timing measurements

[image] saved '/opt/demo/images/out_kfy_20201030195943.jpg'  (1280x720, 4 channels)

[TRT]    ------------------------------------------------
[TRT]    Timing Report /usr/local/bin/networks/SSD-Mobilenet-v2/ssd_mobilenet_v2_coco.uff
[TRT]    ------------------------------------------------
[TRT]    Pre-Process   CPU   0.07802ms  CUDA   0.48875ms
[TRT]    Network       CPU  45.52254ms  CUDA  44.93750ms
[TRT]    Post-Process  CPU   0.03193ms  CUDA   0.03177ms
[TRT]    Total         CPU  45.63248ms  CUDA  45.45802ms
[TRT]    ------------------------------------------------

[gstreamer] gstCamera -- stopping pipeline, transitioning to GST_STATE_NULL
[gstreamer] gstCamera -- pipeline stopped

We are using the enhanced example script, detect.py.   To capture a webc amera image and classify:   camera = jetson.utils.gstCamera(width, height, camera)

This is plenty fast and gives us the results and data we want.





References:

Popular posts from this blog

Ingesting Drone Data From DJII Ryze Tello Drones Part 1 - Setup and Practice

Ingesting Drone Data From DJII Ryze Tello Drones Part 1 - Setup and Practice In Part 1, we will setup our drone, our communication environment, capture the data and do initial analysis. We will eventually grab live video stream for object detection, real-time flight control and real-time data ingest of photos, videos and sensor readings. We will have Apache NiFi react to live situations facing the drone and have it issue flight commands via UDP. In this initial section, we will control the drone with Python which can be triggered by NiFi. Apache NiFi will ingest log data that is stored as CSV files on a NiFi node connected to the drone's WiFi. This will eventually move to a dedicated embedded device running MiniFi. This is a small personal drone with less than 13 minutes of flight time per battery. This is not a commercial drone, but gives you an idea of the what you can do with drones. Drone Live Communications for Sensor Readings and Drone Control You must connect t

Migrating Apache Flume Flows to Apache NiFi: Kafka Source to HDFS / Kudu / File / Hive

Migrating Apache Flume Flows to Apache NiFi: Kafka Source to HDFS / Kudu / File / Hive Article 7 -  https://www.datainmotion.dev/2019/10/migrating-apache-flume-flows-to-apache_9.html Article 6 -  https://www.datainmotion.dev/2019/10/migrating-apache-flume-flows-to-apache_35.html Article 5 -  Article 4 -  https://www.datainmotion.dev/2019/10/migrating-apache-flume-flows-to-apache_8.html Article 3 -  https://www.datainmotion.dev/2019/10/migrating-apache-flume-flows-to-apache_7.html Article 2 -  https://www.datainmotion.dev/2019/10/migrating-apache-flume-flows-to-apache.html Article 1 -  https://www.datainmotion.dev/2019/08/migrating-apache-flume-flows-to-apache.html Source Code:   https://github.com/tspannhw/flume-to-nifi This is one possible simple, fast replacement for " Flafka ". Consume / Publish Kafka And Store to Files, HDFS, Hive 3.1, Kudu Consume Kafka Flow   Merge Records And Store As AVRO or ORC Consume Kafka, Upda

Advanced XML Processing with Apache NiFi 1.9.1

Advanced XML Processing with Apache NiFi 1.9.1 With the latest version of Apache NiFi, you can now directly convert XML to JSON or Apache AVRO, CSV or any other format supported by RecordWriters.   This is a great advancement.  To make it even easier, you don't even need to know the schema before hand.   There is a built-in option to Infer Schema. The results of an RSS (XML) feed converted to JSON and displayed in a slack channel. Besides just RSS feeds, we can grab regular XML data including XML data that is wrapped in a Zip file (or even in a Zipfile in an email, SFTP server or Google Docs). Get the Hourly Weather Observation for the United States Decompress That Zip  Unpack That Zip into Files One ZIP becomes many XML files of data. An example XML record from a NOAA weather station. Converted to JSON Automagically Let's Read Those Records With A Query and Convert the results to JSON Records