Cloud Machine Learning

Ben10 · Post by **Ben10** » Sat Aug 11, 2018 4:21 pm

So I'm training a neural network, through Tensorflow, on the cloud - Cloud Machine Learning.

I tried using Google Cloud Project, Amazon and Paperspace, but both require a credit card to sign up - which I would rather not give.

So does anyone know a cloud machine learning service which I can use to train my model and does not ask for a credit card?

Thanks,

Ben10

Ben10 · Post by **Ben10** » Mon Aug 13, 2018 7:38 pm

Update:
I think that I've found a solution. So today I went down to the local bank and opened a new account, with a debit card - which in 4 days, I will use to sign up.

Liam · Post by **Liam** » Mon Aug 13, 2018 10:50 pm

Good work. Yeah your post showed a valid need for a free cloud software, at least temporarily. Was struggling to come up for a solution for you

L

~ Andrew · Post by **~ Andrew** » Fri Aug 17, 2018 3:13 pm

Latest Mag Pi E-Magazine has several articales on google on a Pi stand alone object recognition and another article on Tensor Flow that uses your video camera to do damn smart stuff...
FREE Download and learn all about it...

https://www.raspberrypi.org/magpi/

~ Andrew

Ben10 · Post by **Ben10** » Sun Aug 19, 2018 8:11 pm

Got my debit card in the mail on Friday. Right now as I type this, I'm training a model on gcloud ml engine - hopefully it goes well

.

Thanks for the link to the magazine - great to see what is possible with open source code.

Thanks.

Michaelcak · Post by **Michaelcak** » Mon Sep 03, 2018 5:53 pm

machine learning is easy to use, but hard to master.

there are plenties to api,third party tools and scripts to do machine learning. it depend on what kind of algo you are looking?

what kind of network? and what kind of learning you want to achieve.

Ben10 · Post by **Ben10** » Thu Sep 06, 2018 7:19 pm

So currently I'm using tensorflow with the Object Detection API - https://github.com/tensorflow/models/tr ... _detection, to train a neural network to draw boxes around objects.

To save time I'm using transfer learning on a COCO-pretrained model. The model I used for transfer learning is ssd_mobilenet_v1_coco_2018_01_28 (http://download.tensorflow.org/models/o ... _28.tar.gz), which I believe means it uses a Single Shot Multi-box Detector (SSD) model (https://arxiv.org/abs/1512.02325), along with the mobile net model https://arxiv.org/abs/1704.04861

~ Andrew · Post by **~ Andrew** » Tue Sep 18, 2018 12:53 pm

I made a picaxe LEARN the position of a servo and then play it back...
1/. Measure the current to the servo
2/. Monitor the current while feeding the servo feeble 'feeling' pulse rates of control
3/. SENSE the current versus the signal being fed to the servo as it is gently 'forced'
4/. REMEMBER the position desired (controlled by current dip at correct position)
5/. Play it back in a loop

Kind of fun circuit with essentially just a 1 ohm resistor as the sensor and some manual servo sense and control loops. With picaxe you can build all these ideas up from absolute basics.
~ Andrew

manisha123 · Post by **manisha123** » Sun Dec 09, 2018 1:40 am

I am also taking training in Machine learning from < a href =" http://www.cetpainfotech.com/technology ... e-learning" /> Machine learning training in Noida </a>

Ben10 · Post by **Ben10** » Sun Dec 09, 2018 2:50 pm

That's great, but Udacity does have some pretty good courses too, that have been sponsored from the likes of Google, AWS and Nividia.

Example course by Google on Udacity (IT'S FREE TOO):
https://www.udacity.com/course/deep-learning--ud730

The Forum