How To Use DeepCognition To Build Drag And Drop Deep Learning Models Without Coding?


Deep studying is an integral a part of synthetic intelligence and the contributions performed within the discipline is immense. With growing analysis and improvement in deep studying, there was a rise in using no-code platforms for deep studying as nicely. There are a number of platforms that help machine studying and the processes like knowledge visualization, processing and so forth. But there are few platforms that focus solely on deep studying and one such platform is DeepCognition. 

In this text, we’ll study somewhat bit about DeepCognition and construct an algorithm utilizing DeepCognition platform. 

Who are

DeepCognition was based with an intention of democratization of artificial intelligence. They have created a platform that can be utilized to create and deploy deep studying fashions with simply clicking of buttons and no code in any respect. 

The downside they’re making an attempt to resolve is to beat the scarcity of experience in AI that’s creating boundaries in organizations within the adoption of AI and make deep studying accessible to all.

Features of

Some of the helpful options supplied by this platform are:

  1. Designing, visualizing and coaching of deep studying fashions with completely no code. 
  2. Provide superior pre-trained fashions like masks RCNN, DenseNet, MobileNet or construct your personal customized mannequin. 
  3. Improve safety in AI
  4. Supports AutoML for hyperparameter tuning.
  5. Easy use of the platform by home windows and ubuntu customers. 

Installation of DeepCognition

To start working with the platform we have to set up it first. Head over to this hyperlink and create an account first. The platform is freed from value for all customers. After creating your account you will notice this web page. 


Based on the working system you have got both choose ubuntu or home windows. Once you click on on the icon the obtain will mechanically start. It will take a couple of minutes to obtain the dependencies and for the platform to open. 

After the obtain is full, click on on the DeepCognition icon on the desktop and you will notice this field. 


Here, kind in ./dlsctl begin as proven above and the platform will open after checking the out there ports. 

After the platform opens you may see this web page.

Exploring a dataset

There are frequent public datasets out there within the platform itself like MNIST, titanic and so forth. you may see this by clicking on the datasets choice on the right-hand facet. 


You can both use the general public datasets or you may add your personal datasets by clicking on my datasets choice. Here, you will notice an choice to add the datasets out of your system.

Since this can be a tutorial on the platform I can be utilizing the general public dataset. I’ll make use of CIFAR10 dataset for the implementation. 

Create a undertaking

The first step right here is to click on on Project-> New and provides a reputation and outline of your undertaking. 

After this click on, the inexperienced arrow and your undertaking will get created. After this, you’ll click on in your undertaking and might see the dashboard which has a number of choices. 

Here as you may see the dataset is chosen. You can change the practice and take a look at break up and the choice to load the dataset both one batch at a time or all collectively. After this click on, the Next button and you will notice a web page to create your deep studying mannequin. 

Model constructing

As you see above, there are a number of choices so that you can select from. You can first choose the enter layer, then convolutional layers, max-pooling and even core layers like flatten and dense. This is nice while you need to construct customized fashions. But to make use of the true energy of the software you’ll click on on the final icon within the bar which is able to pop up the AutoML choice. 

This will mechanically design a neural community for you while you specify the kind of enter and the kind of output. Then, click on on the design button and you will notice your entire mannequin constructed for you. 


The superb half right here is that as and while you construct the mannequin the code is mechanically generated for you. 


So, you’ll not solely get the mannequin but in addition the code to your personal use. 

Hyperparameter tuning

The subsequent half earlier than coaching is tuning the hyperparameters. Again, this occurs with a lot ease within the platform. 


As you may see it is possible for you to to set the epochs, batch dimension, optimizer and the loss. I’ve made modifications to go well with my program as follows. 


After constructing the mannequin and setting the parameters as wanted, we will transfer to the coaching half. 

To practice the mannequin you first want to pick out the CPU or GPU that you’ll use. These options can be found at no cost just for a restricted time and you may choose whichever is suited to the undertaking. 

I’ve chosen the four GB GPU. Once performed, click on on begin coaching and the coaching begins. You can see persevering with modifications on the display screen as and when the coaching proceeds. Depending on the GPU the time taken for completion will range. 

Once I click on the beginning button, you may see the modifications under. 

This course of will proceed till all epochs have been accomplished and the weights of the mannequin are mechanically saved. 

Once the coaching is accomplished you will notice the ultimate accuracy rating and the loss. 

This appears to be a superb accuracy for 10 epochs however we will get higher outcomes with extra tuning of the mannequin. 


After coaching the mannequin we have to see how nicely the mannequin carried out and if there’s any overfitting that has occurred. 

To do that click on the consequence part and you may see the graphs on this part.


You can see that there isn’t any sudden spike and there’s no overfitting of the mannequin. The last a part of the undertaking is the inference of the mannequin. 


After coaching and getting the outcomes we’ll test the predictions that the mannequin has performed up to now and acquire some inference. To do that click on the inference part and you may choose the mannequin weights and click on the beginning inference button.


Once the inference is accomplished after a couple of minutes the predictions seem on the display screen. 

Some of the predictions are unsuitable however there are extra proper predictions as nicely. Not solely this, even the possibilities of the predictions which are the output from the softmax perform can also be proven right here. 

You can hold these weights and deploy your undertaking from the platform as nicely. 


In this text, we noticed an introduction to and step-by-step constructed a whole deep studying mannequin with out having to write down even one line of code. This is of nice use for non-programmers and enterprise analysts to work on deep studying initiatives with ease and at no cost of value. 

If you liked this story, do be a part of our Telegram Community.

Also, you may write for us and be one of many 500+ consultants who’ve contributed tales at AIM. Share your nominations here.


Please enter your comment!
Please enter your name here