Please visit the following URL and read about a few things you should consider doing to prepare and take full advantage of the workshop:

https://bit.ly/2TqZ6CA

Visual AI solutions combined with powerful sound diagnostics for real-time decision-making are some of the hallmarks of the Sony Spresense with Edge Impulse’s embedded ML technology. Together, Edge Impulse and Sony bring a unique combination of solid computing performance as well as serious power efficiency that is ideal for edge computing applications. 

Join our hands-on workshop to learn how to build future-proof solutions with smart sensor analysis, image processing, data filtering, collecting raw data, getting insight into that data using signal processing and machine learning, and deploying your ML models, ready for scale and industrial production.

• Learn how embedded ML gives real-time insights into complex sensor streams
• Build your first embedded ML model in real-time
• Gain insight into the types of problems ML solves, then build better products
• Learn how to take your ideas to production and scale through complete MLOps

Workshop details:

• A 90-minute workshop
• Beginner/Intermediate skill level
• Hands-On, Instructor-Led, Live
• A recording will be shared post-event
• A personalized Certificate of Accomplishment from Edge Impulse
• Purchase your Sony's Spresense kit from Adafruit today or check here for more buying options.

15:13:15	 From  Justin Lutz : are current edge detection mcu's capable of doing object detection?  I thought we were limited to just image classification?
15:14:02	 From  Aditya Mangalampalli (Edge Impulse) : Yup, Edge Impulse boards are fully functional for both Image Classification and Object Detection among lots of other neural network architectures!
15:14:59	 From  Justin Lutz : ok, thanks!
15:15:08	 From  Andrew Siemer (Edge Impulse) : We have some great examples for object detection using the Raspberry Pi and Jetson Nano.
15:16:16	 From  Justin Lutz : yes, I know we can do that on the SBCs, didn't think MCUs were powerful enough for that yet
15:17:08	 From  Aditya Mangalampalli (Edge Impulse) : Yup, with Edge Impulse’s custom compiler, MCUs are now able to run the same things as some SBCs are!
15:17:35	 From  Justin Lutz : wow, awesome, can't wait to try it out
15:17:59	 From  Noel Putaansuu : Will it work with the esp32?
15:19:14	 From  Zin : We are looking at providing object detection for MCU class processor with something like an Cortex-M7 being the first target
15:19:53	 From  Bukasa Tshilombo   to   Stephane Boucher(Direct Message) : Thanks. I got in. Bukasa
15:20:36	 From  Vishnu B Raj : Can we use our python codes in edge impulse and optimize it for spresence?
15:20:37	 From  Rahman, Mehdi : Are there label annotation tools in the Edge Impulse interface?
15:21:09	 From  Aditya Mangalampalli (Edge Impulse) : There are label annotation tools in the Edge Impulse Studio, that is correct.
15:21:31	 From  Zin : We are seeing MCUs add more and more sophistication (e.g. NN accelerators), narrowing the gap between MCU and SBC, of which we’ll be able to leverage
15:22:10	 From  Justin Lutz : @Zin, makes sense thanks
15:25:23	 From  Muhamed Fauzi Bin Abbas : Where can we find the C-library for inferencing?
15:25:28	 From  Zin : @Vishnu B Raj you can define a more detailed custom NN architecture in python using “expert mode” on the training page
15:25:59	 From  Vishnu B Raj : @ZIn Cool Thanks
15:26:01	 From  Justin Lutz : I've taken the course and I recommend it as well!
15:26:10	 From  Rahman, Mehdi : Where does the training happen?  On cloud machines?
15:26:21	 From  LDC : Do you have a link to the course?
15:26:32	 From  Anubhav Agarwal (Edge Impulse) : @Rahman Yes, training happens in the cloud.
15:26:43	 From  Jenny Plunkett (Edge Impulse) : https://www.coursera.org/learn/introduction-to-embedded-machine-learning
15:26:54	 From  LDC : Thank you
15:27:52	 From  Zin : @Muhamed Fauzi Bin Abbas you will get this when you export to c++ library on deployment page (after you finish training a model)
15:28:03	 From  Rahman, Mehdi : So my understanding is that there is a cost per hour type of system for that, correct?  We could also train using our own system and then serialize the model to run on a controller?  Let's say we already had a tensorflow or pytorch model?
15:28:05	 From  Jenny Plunkett (Edge Impulse) : https://studio.edgeimpulse.com/public/37001/latest
15:28:25	 From  Dave Nadler : Can you clarify training before deployment and edge training? So far you seem to be talking just about running inference at the edge? Thanks!
15:35:52	 From  Andrei : How do we connect the device to a different project... like the one we just cloned?
15:36:22	 From  Jenny Plunkett (Edge Impulse) : edge-impulse-daemon --clean
15:36:29	 From  Andrei : Thanks, I tried with just one dash
15:36:39	 From  LDC : Where do you type that?
15:36:50	 From  Justin Lutz : I just had to do the same thing as well after my testing yesterday :)
15:37:14	 From  Andrew Siemer (Edge Impulse) : https://docs.edgeimpulse.com/docs/cli-installation#installation---macos-and-windows
15:37:27	 From  Jenny Plunkett (Edge Impulse) : https://edgeimpulse.notion.site/Edge-Impulse-Embedded-Online-Conference-Participant-Pre-requisites-July-27-2021-d3226bc26f374794b0bb554096d95424
15:39:22	 From  LDC : Any quick hint at that stage if you get an access denied error for the com port when trying to connect to the device (on Windows through PowerShell)
15:39:49	 From  LDC : I did that already
15:39:51	 From  LDC : same issue
15:40:11	 From  Jenny Plunkett (Edge Impulse) : Start-Process powershell.exe -ArgumentList ("-NoExit",("cd {0}" -f (Get-Location).path)) -Verb RunAs
15:50:26	 From  Rahman, Mehdi : What kinds of data augmentations does Edge Impulse allow?  Flips, rotations, crops?
15:51:38	 From  Vishnu B Raj : That's awesome
15:53:03	 From  Rahman, Mehdi : Is there a cost to training the model on the cloud?
15:53:27	 From  LDC : (I made progress connecting using your link- now I get "Serial is connected, trying to read config...)
15:53:29	 From  Aditya Mangalampalli (Edge Impulse) : @Rahman, we support all augmentations that are included within the Tensorflow Augmentation module (exposure, flips, crops, etc.)
15:53:30	 From  Aditya Mangalampalli (Edge Impulse) : https://www.tensorflow.org/tutorials/images/data_augmentation
15:53:41	 From  LDC : (and repeated timeouts)
15:54:04	 From  LDC : Many
15:54:07	 From  Justin Lutz : do you have fw on spresense board?
15:54:08	 From  LDC : (continuous)
15:54:16	 From  Rahman, Mehdi : Thanks @Aditya
15:54:33	 From  Vishnu B Raj : Does hyperparameter tuning done here automatically in case of custom models?
15:54:34	 From  LDC : (Yes I did....and we can likely solve this later unless others are having trouble)
15:54:41	 From  Jenny Plunkett (Edge Impulse) : https://docs.edgeimpulse.com/docs/sony-spresense#3-update-the-bootloader-and-the-firmware
15:54:43	 From  LDC : I will retry that
15:54:43	 From  Anubhav Agarwal (Edge Impulse) : @Rahman, Mehdi For audio data we have other types of augmentation such as axis warping, noise, etc.
15:55:11	 From  Justin Lutz : might need new bootloader as well
15:55:18	 From  LDC : Got it . Thanks
15:55:19	 From  Justin Lutz : I had to do that via Arduino IDE
15:55:37	 From  Darryl : Previously I tried flashing both this one and also the one below (Arduino).  Did not seem to work.  I also had to modify the baudrate.
15:56:22	 From  David DeFilippis : @Justin Lutz, I ended up having to do the same thing. I couldn’t get the EI bootloader working on the Spresense until I loaded the Arduino bootloader first.
15:56:23	 From  Anubhav Agarwal (Edge Impulse) : @Rahman, Mehdi you can also extend the augmentations in Expert mode easily
15:56:23	 From  Darryl : I think the Arduino is on the same page just lower
15:56:39	 From  Justin Lutz : @David, yup
15:56:42	 From  armaghan : For Arduino please use this: https://developer.sony.com/develop/spresense/docs/arduino_set_up_en.html
16:04:44	 From  Vishnu B Raj : There might be more than 3 features right? How are these 3 features selected in this visualization?
16:04:54	 From  Justin Lutz : if you are using your own dataset, is there a quick way to split the data between training and test?
16:05:24	 From  Jenny Plunkett (Edge Impulse) : https://umap-learn.readthedocs.io/en/latest/
16:05:27	 From  Vishnu B Raj : Thanks
16:05:38	 From  Rahman, Mehdi : How long does the model inference take on the device?  Is it 5 ms passing the image to the model and getting an output from the microcontroller?
16:07:02	 From  Rahman, Mehdi : Does the Edge Impulse interface show you the full timing?  It only shows 5 ms but nothing on passing the features to the model and getting an output?
16:07:32	 From  Justin Lutz : great, thanks!
16:08:38	 From  Rahman, Mehdi : @Jenny, thanks for the clarification!
16:11:30	 From  Vishnu B Raj : In realtime projects is it possible to use data collected from spresence and train again in cloud and update the model automatically?
16:12:23	 From  Rahman, Mehdi : Out of curiosity, what kind of speed boost does the EON optimizer give compared to the base inference without the optimization?
16:13:26	 From  Dave Nadler : Can you do any learning or model adjustment at the edge only?
16:17:42	 From  Justin Lutz : yikes, I got "excessive protocols" error
16:18:00	 From  Justin Lutz : while flashing
16:18:11	 From  Darryl : I had to lower the baud rate to fix that error
16:19:08	 From  Darryl : Mine was way higher to begin.  It worked at 115k
16:19:09	 From  Justin Lutz : thanks will give it a try
16:21:26	 From  Rahman, Mehdi : What is the 391 on the current serial output?
16:21:45	 From  Darryl : I was never able to get beyond the nuttx or Arduino boot prompts.  How is the best way to get help?
16:21:53	 From  LDC : Thank you! Is there a contact link for further debugging if the re-flashing solution doesn't work?
16:22:03	 From  Jenny Plunkett (Edge Impulse) : hello@edgeimpulse.com
16:22:04	 From  LDC : Thank you
16:22:09	 From  Jenny Plunkett (Edge Impulse) : https://forum.edgeimpulse.com/
16:22:36	 From  Justin Lutz : baud rate was at 921k so lower baud rate worke, thanks so much
16:22:51	 From  Jenny Plunkett (Edge Impulse) : https://studio.edgeimpulse.com/public/37001/latest
https://studio.edgeimpulse.com/public/20687/latest
https://studio.edgeimpulse.com/public/20202/latest
https://studio.edgeimpulse.com/public/27835/latest
16:23:48	 From  Jenny Plunkett (Edge Impulse) : https://imaging.framos.com/webinar/spresense-microcontrollerwithnuttx/
16:24:02	 From  Muhamed Fauzi Bin Abbas : How do I incorporate the inferencing library into my own code?
16:24:13	 From  Paul : How do you include your own code?
16:24:19	 From  Paul : Lol
16:24:23	 From  Jenny Plunkett (Edge Impulse) : https://docs.edgeimpulse.com/docs/running-your-impulse-locally-1
16:24:40	 From  armaghan : Thank you Jenny!
16:24:44	 From  Jenny Plunkett (Edge Impulse) : https://github.com/edgeimpulse/firmware-sony-spresense
16:24:48	 From  Jenny Plunkett (Edge Impulse) : https://github.com/edgeimpulse/example-standalone-inferencing-spresense
16:26:33	 From  Rahman, Mehdi : What was the classification=391ms on the serial output earlier?
16:27:40	 From  Rahman, Mehdi : @Jenny, thanks for the explanation
16:28:00	 From  Justin Lutz : great workshop, thanks a bunch!
16:28:08	 From  Paul : Thank Jenny.
16:28:10	 From  Vignesh baskaran : Indeed!!
16:28:17	 From  Priyesh Sharma : This was very informative.. thanks guys..
16:28:27	 From  Adrian Wahl : Wonderful, thank you very much. Lots of material to pick up. great
16:28:43	 From  Noel Putaansuu : Very good thank you.
16:29:03	 From  Justin Lutz : can you post a link here?
16:29:05	 From  armaghan : Great presentation and workshop! Thanks Jenny!
16:29:14	 From  Vishnu B Raj : Just curious!! what's edge impulse's business model?
16:29:20	 From  Harald : Thank you.  Great Workshop.
16:29:38	 From  Jenny Plunkett (Edge Impulse) : For business model questions please email us at hello@edgeimpulse.com
16:29:44	 From  Jenny Plunkett (Edge Impulse) : @Justin link to what
16:30:02	 From  Justin Lutz : the embedded systems/ml workshop in October?
16:30:07	 From  LDC : Thank you
16:30:10	 From  Justin Lutz : that Jacob was talking about
16:30:14	 From  Shahin Etemadzadeh : Thanks, that was fun
16:30:27	 From  Justin Lutz : thanks
16:31:03	 From  Justin Lutz : cool, thanks
16:31:10	 From  Jenny Plunkett (Edge Impulse) : Thank you everyone for joining!
16:31:29	 From  Vishnu B Raj : Thanks
16:31:31	 From  javi : thanks
16:31:36	 From  Paul : Cheers