How to provide stride and Pad


#1

Hi , how i can choose STRIDE and PAD for Conv and Maxpool layers in GUI model environment? there i didnt saw any option for it.

J.Dud


#2

1

there are no option for above mentioned options to choose for conv and maxpool layers. plz someone guide me.

thank you


#3

Hey
Just click on the show advanced option where you will find Subsample which is also known as stride and you can use border mode for keeping the image size same.
Thank you


#4

thank you for helpful reply.
i want to mention some previously DNN structures in the Literature, they also assign Filter size to MaxPool layer, which are not available in DLS, also previous structures mentioned CONV layers STRIDE and PAD, so these two kind of options i cannot find in DLS or may be it availble with any alternative names.
Kindly educate.
Regards
J.Dud


#5

Hey can you send us the structure summary that you are talking about. I will create a DLS model for that using all those features so that you can understand all this better else I will soon send you an article explaining all these features use during model building using DLS.
Regards
Rajat


#6

You can have look to this model (send you through your email). can you please explain that how to design it in DLS, especially what i have ask in my previous query.
Regards
J.Dud


#7

Sure I will soon send you the configuration.yaml file which you can directly upload into DLS.
and for your previous query I have told you the alternative of stride and pad and also I am writing a detailed article explaining all the features of DLS layers.
Regards
Rajat


#8

Hey I can’t find your mail. On which email id you have send it.
Can you mail me that model @ hello@deepcognition.ai .
Regards
Rajat


#9

i have sent the model architecture to your email @hello@deepcognition.ai. how we can design it in DLS?

Best Regards
J.Dud


#10

Yes I have received your mail and as i told you earlier that if you want to use stride in convolution layer then you have to use Subsample in DLS that is available in advance option in convolution layer.

I am attaching configuration file link of the model that you have send for your refrence.
https://drive.google.com/file/d/1udP4ZXxyRm2ml0YVfGBlyZuxrwv2KC5p/view?usp=sharing

To use this file just create a new project
Go to the model tab and click on the upload button as shown below


upload this yaml file and your model is ready to use in DLS.

Regards
Rajat


#11

Thank you for detailed and very helpful inputs. :grinning:
here i want to ask about the output layer,no need to put output layer or just the last layer you mentioned with softmax is enough for output?

Regards


#12

Welcome :smile: and Yes you have to put the output layer in the end but i didn’t because i thought you might want to extend this model since it still require flatten layer and dense layer for multi-class classification and after which you will add the output layer in the end.

Regards
Rajat