Pytorch Scholarship Challenge

Source: Deep Learning on Medium


#pytorchudacityscholar

Earlier on 30th October 2018,i received an email congratulating me for having been accepted to Udacity’s PyTorch Scholarship Challenge by Facebook.

I was excited that i long waited for the classes to begin.Finally,the day came and i was directed to join my classmates,mentors and project managers in slack community.During the program,we got a chance to help each others out,learn amazing things.

The Challenges I Faced

It was not a walk in the park given that i was being introduced to deep learning for the first time.I hadn’t written code in python,by then i was only an android developer.To be sincere,i loved python syntax,its flexibility and a lot that it can do with just few lines of code.I took a lot of time learning its basics and some of its powerful modules that is numpy,pandas,sci-py and matplotlib(I loved what this module does,data visualization).I also had to take a data science course from edx. I grasp a lot from the data science course,more importantly understanding the Unix command-line. All of these challenge made me grow,be more flexible and more dynamic.

Slack Community

I don’t know where to begin because there is much to tell about how amazing pytorch slack community is.Any issues was perfectly addressed,different kinds of problems solved regarding the course work.The study jam’s(Learning for 24 hours)Helped a lot in finishing the challenge.Getting help from student leaders,helping and asking questions was an awesome experience.I learnt a lot from the community(working in a team).

What I learnt as a mentor

I got an opportunity to work as a mentor.I was partnered with three scholars to help them finish the challenge in time(@Ayanlola,@risper bevalyn and @fichoch).I have learnt a lot from them as we learnt together,helped each other out.

Final challenge(Image Classifier Model)

The community contributed a lot for my successive submission of the final project.Along side that i had to google a lot.There were so many documentations to help in very many parts of the project.But the most significant ones were the course materials.They greatly helped.Surprisingly,i had just finished the model but faced a challenge on how i uploaded the model to Udacity work-space for only the checkpoint is required.I took four days trying to figure out what what was wrong.This was a big frustration to me.But i came to realize that the problem was how i uploaded the checkpoint file to the GDrive.

wrongly uploading the checkpoint to GDrive

Correctly uploading the checkpoint to GDrive

Model Accuracy.

How the model classified a test image

What I will do with the knowledge learnt

I will integrate the concept learnt to designing models that will fit my personal projects.Participate in Open Source projects that need automation of intelligent behavior(I will apply for Google Summer Of Code).Helping in my programming communities that is Pytorch community pytorch,reddit,stackoverflow etc.This will further increase my skills as i look forward on witting a research paper based on this skills.

Gratitude

Thanks to Facebook and Udacity for giving me an opportunity to learn and add to my skill set.

#pytorchudacityscholar