KCDC: I failed.

Amazing Conference

First, I want to say how amazing KCDC was this week. From the hotel, the venue, the people, and the crowd were all top notch. I will measure the rest of the conferences against this one. The group that puts this on needs all of the praise.

Setting Up For Failure

Now, on to the point of this post. I bombed. I created a new presentation for this conference where I was going to focus on the deployment part of machine learning. I knew there are concerns with each extra external source you need to have work during a live presentation. I had 4. The first was a web site I could host in VS Code using Live Server. But, the next three all could cause major concerns. The Android application needed Android Studio and an emulator. Google Cloud required a billable project as well as a working AI platform. Azure requires a resource group, storage, and the machine learning studio. Way too many moving parts.

My Failure

I had a working presentation on Friday but on Saturday I tried to add a working Android application. This wasn’t smart. I am using my oldest’s computer. It isn’t a development machine. I shouldn’t have tried to make it one. Once I booted up Android Studio and started the emulator the laptop was toast. Presentation over.

The flow of my presentation SHOULD have been creating a Celsius to Fahrenheit TensorFlow model, then taking that and converting the model to TensorFlow.JS and put it in a web site. I would then convert the model to TensorFlow Lite and put it in an Android application. Then, I would take the full model and place it in Google clouds AI engine. Finally, I would place it in Azure.

Since Android Studio made the machine almost worthless I decided to do the Android part first and then close Android Studio. This made my presentation out of order with absolutely zero flow. But, at the moment, I didn’t know what else I could do.

After I got control of my computer back 30+ minutes into the 60 minute talk I was left with trying to cover as much as possible.

Changes

If I had to do this again (and I will at a future conference) I would skip the Android Studio and just show code. Running the demo isn’t worth it with my current hardware.

I also need to figure out how I want to handle the two clouds. I think that I will make those steps screenshots and just show the end result. I need to take some of the variance out of my talks.

Where to go now?

Well, I guess I learn from my mistakes and clean up the project. Add the screenshots to ensure a solid flow to the presentation. Then, only go to the working final solutions at the end of the project if I have time. I also need to dig into how I want to handle both of the clouds so I am not getting billed for all the days leading up to the presentation.

I am also planning to create a YouTube video of the presentation. That will let me do some editing as well as get into a flow.

Conclusion

I fell flat on my face at my largest conference yet. Now, I need to make some changes and move on. I need to ensure that my next presentation is great and not a waste of time for the people in the crowd that are giving me their time.

Links

My presentation slides: Github/ehennis
My presentation content: Github/ehennis

GDG San Diego Talk

Overview

This will be the fourth time I am giving this talk about my work on the Raspberry Pi. I last gave it at the Google Developers ML Summit and it was recording and that was kind of odd. We will see if this one is a little better with some audience feedback.

TensorFlow Lite

Since I had some extra time I was able to implement TensorFlow Lite on the Raspberry Pi. I went to the TF Quickstart [Link] page and since I had already converted the Keras model into a tflite model I only needed the interpreter. This allowed me to not have to have the full TF install*.

*I still used the Keras preprocessing and decoding methods that came with the MobileNetV2 model. I probably should have not done that but at this point I don’t think it matters as much. Maybe in the future I will do ONLY TF Lite.

Links

Original Blog Series: Here
Meetup: https://gdg.community.dev/events/details/google-gdg-san-diego-presents-tensorflow-lite-on-the-raspberry-pi/
Code: Main, MobileNetV2Base, and PiCameraManager
Presentation: ImageDetection/GDG-SanDiego.pptx

Heartland Developers Conference Talk

Overview

This is a version of my talk about adding some image detection onto the Raspberry Pi. For the Paris talk, I only had 20 minutes and for GDG Denver I had 90 but didn’t use all of it so this length should be about right.

TensorFlow Lite

Since I had some extra time I was able to implement TensorFlow Lite on the Raspberry Pi. I went to the TF Quickstart [Link] page and since I had already converted the Keras model into a tflite model I only needed the interpreter. This allowed me to not have to have the full TF install*.

*I still used the Keras preprocessing and decoding methods that came with the MobileNetV2 model. I probably should have not done that but at this point I don’t think it matters as much. Maybe in the future I will do ONLY TF Lite.

Links

Original Blog Series: Here
Meetup: https://www.heartlanddc.com/
Code: Main, MobileNetV2Base, and PiCameraManager
Presentation: ImageDetection/HDC.pptx

Iowa State Google Developer Student Club Presentation

I have tried to get a student club created at Iowa State ever since I became a GDE. Well, the people at Google with some real pull were able to find some great students to lead it. Well, I am going to give a presentation on 9/9 about my work with deep learning and trying to predict basketball scores. Hopefully, they will get some entertainment and learn something.

DSC Link: https://dsc.community.dev/events/details/developer-student-clubs-iowa-state-university-presents-machine-learning-workshop-with-evan-hennis-google-developer-expert/

Presentation: https://github.com/ehennis/Blog/blob/master/DeepLearningBasketball/DL-IowaState.pptx