Anatomy of Deep Learning Frameworks

Anatomy of Deep Learning Frameworks

Gokula Krishnan Santhanam

Gokula Krishnan Santhanam

Zürich, Zürich

Everything you wanted to know about deep learning frameworks

Artificial Intelligence

  • 0 Collaborators

  • 0 Followers

    Follow

Description

Blog post and video explaining how various deep learning frameworks and how you create your own. Done as a part of the Indian Deep Learning Initiative webinar series.

Video

Links

Medium link to the blogpost

Medium whatsapp image 2017 08 21 at 11.51.22 pm

Srihari J. created project Novel approach to avoid glare while driving cars in Highways at night.

Medium c96d72d5 eb8c 4f98 afce 405b7a700812

Novel approach to avoid glare while driving cars in Highways at night.

Abstract:APT i.e "Accident Prevention Technology".All we see in the second page of any News paper is about unfortunate incidents ,The Accidents.We are coming up with amazing features for all four wheelers which on support with statistical reports, prevent 50% of the accidents. That's the minimum margin we put, as our Arduino, Ultrasonic and Servo-motor based quick responsive systems, and android map application, taking care of the obstructions caused for the drivers which leads to accidents(glare hitting the eyes when other vehicles zoom in with high beams "on" highways in night), and about the junctions, which are very difficult to spot when you are driving at 40+mph and none to warn you about the nearest intersection(our android application using G-Maps API takes care of it)!! Aren't intersections ,glare, and drowsiness main reasons for accidents?! We are taking at most care to address each and every aspect. We prevent the glare hitting the drivers eyes (completely automated systems , implementable for every lower end/economical cars),using Robotic hands, which is guided by an image processing system along with Light Detecting Resistors value. Also keep the vehicles speed in check, mostly at junctions. We will also want to take care about the "Care" part of our project. We make sure the emergency wheels reach out soon to the hospitals, using IoT and GPS system! We are automating the communication system, using the best of the resources available. We are using cost effective sensors, and connecting each and every device, bringing out a working prototype. Our total expenses comes under a thousand rupees , for the completely working and ready to implement systems. We feel this is the perfect platform to bring out our innovations and make them working models, and reach the world. We await anxiously to execute/implement our thoughts and "SAVE LIVES!!!".

Objectives: Preventing accidents and saving lives, by preventing the glare of opposite direction travelling vehicle disturb the driver. Warning about the very next intersection/junction when the drivers speed is above the limit for that specific road. Automated signal systems which create green corridor for emergency vehicles.

Outcomes: Reduction in accident rate by 50%.

The overall block diagram :

Brief description of the Methodology: a) Glare prevention system: We have used a small camera, fixed on the dashboard, which clicks the picture of the driver every second (automated) and sends the result to the servo motor via Arduino board. Using matlab , we have coded for drowsiness and glare detection. If the output gives to be drowsy, an alarm (beep) rings. If the glare is detected, servo motor is given the right value for the amount of inclination to obstruct the light from falling on drivers eyes. We have used X-ray sheet as a translucent material which obstructs the light in one way direction. The communication between different devices is automated and wireless. b) Junction alarms: If the vehicle is travelling more than prescribed speed for that highway/road ,our android application shouts/alarms that there a junction ahead of 600m and advices the driver to slow down. c) Emergency vehicle systems: Pure IoT concepts, to inform the cop about its arrival to the next signal. One application which also communicates with the signal system, hence creating green corridor and pushing the chances of saving lives.

Expected Results: The human kind is the end user! Ideas themselves call to be philanthropic. None wishes to end his/her life on roads. Everyone cannot afford for automated systems which come in Benz/BMW's. Our glare prevention technology can be installed in less than 2k rupees. The junction app is free and will always be. Ambulance system app again is free, but interacting with the government officials will do the required job for us!

Conclusion: This is an essential implementation which should not be ignored at all. Statistics says 50% of accidents happen summing up of intersections and glare. We single handedly prevent both. We await to implement it at the very best platform we have, ABB Makeathon.

Future Scope :Research and development takes things to next level. Our glare prevention system, having mechanical constraints, work in less than 1.2 seconds. Higher level of automation and accuracy can be confronted. This kind of a concept does not exist anywhere till date.

Medium me car

Devinder K. created project Understanding the Decision Making Process of Deep Neural Networks

Medium 2c89d1fc d576 4086 8d82 85a32832a6ea

Understanding the Decision Making Process of Deep Neural Networks

In this work, we propose CLass-Enhanced Attentive Response (CLEAR): an approach to visualize and understand the decisions made by deep neural networks (DNNs) given a specific input. CLEAR facilitates the visualization of attentive regions and levels of interest of DNNs during the decision-making process. It also enables the visualization of the most dominant classes associated with these attentive regions of interest. As such, CLEAR can mitigate some of the shortcomings of heatmap-based methods associated with decision ambiguity, and allows for better insights into the decision-making process of DNNs. Quantitative and qualitative experiments across three different datasets demonstrate the efficacy of CLEAR for gaining a better understanding of the inner workings of DNNs during the decision-making process.

Default user avatar 57012e2942

KUSHIKA A. updated status

Default user avatar 57012e2942

KUSHIKA AGARWAL

Hello guys !! I am working on Lung cancer detection using CT scans of patient and using DICOM. This involves concept of machine learning and big data.

Medium kron228888

Pinar K. created project Apples or Oranges

Medium 2d5024de 7b24 4292 a121 5ab9615c6d9b

Apples or Oranges

For quite some time now, I have been interested in machine learning. Although I completed a machine learning project for my Bachelor's Thesis which was in MATLAB, I really want to have some hands on experience in Python, too.

So I went ahead and searched for some introductory courses online and I've found Mr. Josh Gordon's awesome video "Hello World - Machine Learning Recipes #1" on Youtube (

). It's a supervised learning project which creates a classifier from examples.

You can access sourcecode in my repository on Github -> https://github.com/pinarkaymaz6/python-supervised-learning

For the absolute beginners like myself, here's some tips to install Python and reqired libraries.

  1. Download Anaconda from here: https://www.continuum.io/downloads I downloaded Python 3.6 version.

  2. Add Anaconda into Path. Control Panel\System and Security\System > Advanced System Settings > Advanced Tab > Environment Variables> System Variables

Find "Path" variable and add Anaconda3 file path. For me, it was "C:\Users\PINAR\Anaconda3"

  1. Open Command Prompt to check Python is successfully installed. Type "python". If you can see the version, everything is okay so far. You can close this terminal now.

C:\Users\PINAR>python Python 3.6.1 |Anaconda 4.4.0 (32-bit)| (default, May 11 2017, 14:16:49) [MSC v.1900 32 bit (Intel)] on win32 Type "help", "copyright", "credits" or "license" for more information.

  1. Create your Python project and copy the source code. It's a simple text file with .py extention. Or, you can download sourcecode from my Github profile.

  2. Open Command Prompt and run your code. Congrats!

python C:\Users\PINAR\Anaconda3\applesororanges.py

Now, I want to develop this project adding statistical analysis. I want to analyse in what ratio this machine can successfully detect apples or oranges. Also I want to add some features to improve these detection results.

Anyone who wants to contribute is more than welcome!

Pınar

Medium 18622237 1164276773718635 1801559994912150109 n

Anurag S. updated status

Medium 18622237 1164276773718635 1801559994912150109 n

Anurag Sen

Just thought about a software which will automatically show up the used time of electronic gadget when it is switched off

See More

No users to show at the moment.

No users to show at the moment.