[RFC] TensorFlow Deep Learning Software Coding

Japanese Farmer using deep learning to automate Cucumber grading, does anyone at DMS familiar with this?

2 Likes

I think it’d be fun to do some sort of machine learning-related event at some point. I went through the process of running CuDNN on a system with my nVidia video card, but I don’t want everyone to have to lug a desktop with an expensive video card to DMS. :stuck_out_tongue: TensorFlow would make this a whole lot more practical, but I haven’t messed around with it yet.

1 Like

I’ve played around with tensorflow. If you have time and an internet connection I can get you up and running a vanilla CNN to classify MNIST.

If you want to get more intimate with the basics of deep learning for computer vision, I suggest looking at this course.

http://cs231n.github.io/

2 Likes

Hehe, MNIST is exactly what I did with my CUDA cores. I was trying to figure out how to convert the HDF5 output into a format where I could run the same algorithm as part of an Android app so it could classify numbers you draw on the screen with your finger. Seems like TensorFlow would make this sort of thing easier.

Thanks for that link, Otto! Next time I see you and have my laptop, be ready… :wink:

1 Like

GOD BLESS YOU! Yes! Can we meet next week?

If you have time and an internet connection I can get you up and running a vanilla CNN to classify MNIST.

@Josh_Melnick Yes we can over hangouts or Skype. Unfortunately I’m no longer Dallas enough for the DallasMakerSpace. I’ve made a bold move to the east coast two weeks ago. Meeting in person will be tough.However, I’m glad to see we have other members interested in this!

@mrcity Remember Quantopian? You should check out the network trainer equivalent http://aetros.com/
In Aetros you drag and drop gui elements to build and define hyperparameters. This gets compiled to Keras code and gets efficiently trained on a dataset of your choice. This is cool because you can either upload your own, or clone/access someone else’s.
Members also frequently publish simple image classification models that are available for cloning and peer review. Notable ones are:

2 Likes

May I have your email? Skype?

Hi, are you familiar with Tensor Flow?

So… no… :wink:

If we can find a time to meet within the next few days, we could take a bit of time to explore it together possibly. And there might be others wishing to join. I’m pretty much free all weekend.

Skype is conicrelief
Email is [email protected]

1 Like

@Josh_Melnick @mrcity When you guys get together, could you ping me? I’d like to shoulder-surf your adventure if you don’t mind!

1 Like

I have some stuff for which I am using Torch. Any of you using Twnsorflow have experience transitioning from Torch?

@Josh_Melnick and others, are you still looking for TensorFlow help? I was thinking about hosting a workshop on TensorFlow and maybe one or two other APIs on 11/17 at 7 PM. I’ll put it on the calendar shortly.

2 Likes

I can’t make it that day, is it possible to do it on the weekend?

I don’t see why not, though I already submitted the calendar request and asked for honorarium. If they’ll let me move it, and can communicate that to the GDG (Google Developer Group) people who are interested in helping promote, then we can have it on the weekend. Worst case scenario, we can meet up sometime mutually convenient.

Is this set as-is on the calendar? Or is that even still subject to change?

It’s set as-is on the calendar, and already 49 people have RSVP’d through the GDG Dallas Meetup page. I will more than likely hold another one at a later date, especially if this one gets a long waitlist.

1 Like

For those of you planning to attend, please take a look at these prerequisite setup steps if you would like to follow along on your own machine with your own accounts & API endpoints: https://github.com/mrcity/mlworkshop/ It may get updated between now & Thursday, so stay tuned.

Expect a packed house!

I can’t attend on THURS can we do another weekend? Any reading material to look over?

1 Like

Will we be able to do another one in a week or two, on a Sat or Sunday?