I am Natalija (@cetalingua) and I am interested in all things audio. My focus is primarily marine mammals vocalizations and I have been using fast.ai to identify signals of interest in acoustic recordings. You can check my first, very rudimentary model here. I plan to work on this model more, so it is better at generalizing and dealing with more complicated audio recordings. Eventually I am planing to move to using CNN to categorize signals, and finally using NLP models to determine if whales and dolphins have any sort of sophisticated communication patterns resembling language. I have been trying to learn DL since 2018 and still is a newbie (and I keep asking stupid questions, like how can I upload 10,000 files to GCP and set up a path to them in Lesson 1, if anyone knows, please let me know!)
@jeremy I hope that we might get at least one audio lecture maybe? People did such a great job with fast.ai audio !
fastai2’s audio is still being developed last I checked (which was not too long ago). Once it’s done I know I’ll be trying to include it in the study group
I am Vijay Narayanan. Currently working as a pre-sales specialist in a Software MNC in Mumbai, India. Started my AI journey as a self skilling initiative with the Stanford ML course by Andrew Ng. Have alsocompleted two Udacity Nanodegrees on AI and Deep Learning.
I have been involved with fast.ai live since 2017. My learning has progressed steadily with a lot of “maybe I am not good enough” to “maybe I am starting to get this”. Through fast.ai i have met amazing people, took part in Kaggle competitions (won a silver medal in one of them ), hosted study groups locally and also on TWiMLAI and have also written blog posts on fast.ai forum explaining some of the concepts that i understood.
If I look back at this journey, it is nothing short of a miracle. From not having any coding background to what I am now, this wouldn’t be possible without this course, this community and the forums. Look forward to learning V2 and doing some projects with some of you .
My name is Hao. In 2018 Dec, I started my deep learning journey with zero experience. As you can tell, only been a year into this wonderful world. Now I am interviewing in the Bay Area and transitioning to machine learning engineer roles.
This all happens because of fastai, just following Jeremy’s lecture, reproducing the notebooks line by line. It is not easy, I remember for some lectures, I have listened again, again and again to a moment that I pretty much can tell what the next scripts are.
This is where I met Radek / DrHB and many others Just by reproducing their reports, learning the concept of handling imbalanced classes, building siamese network, testing efficient net models… etc, I got my first kaggle silver medal… first PR to fastai library.
Special thanks to this great community and fastai course. And looking forward to joining the v2 journey!
Wow, lots of working professionals here! I am a student at Indiana University pursuing my Masters. I had done a deep learning course before fastai but didn’t get anything out of it. Fastai has really been life-changing for me. It has put me on track in terms of learning deep learning, searching for a job and starting my own technical blog. The course and the material just stand out for me. Everything else seems boring. Really glad to be selected for this course and looking forward to collaborating and learning from it as much as possible.
I’m Boris Dayma. I’m currently working as an independent consultant for companies such as Weights & Biases and try to bring automation through DL in the industry.
I have more of an Engineering and Scientific background and went on my own towards AI a few years ago with the help of all the online resources & MOOC’s (so glad it became popular around the same time as DL was growing).
I was initially working mainly with Keras & Tensorflow but then switched to Pytorch and loved it (as fun as using pure python). Then I came across fastai and I really enjoyed all the built-in techniques that you can switch on by just using a Callback so I decided to spend more time learning the library.
I have now been playing already quite a lot with fastai2. It was slightly overwhelming at first with all the decorators but I’m starting to get a hang of it! Also I love the community and the quick reactions on the forums. Excited to learn more!
You’re welcome! To clarify, you’ll have to open a new terminal window and enter the gcloud compute scp command locally (rather than in your tunneled connection).
Absolutely - it will happen. Not in this part though, sorry! I want to wait until we can do a really good job of it. Pretty sure the next course will be the one…
Hi,
I am Ajaykumaar, I am doing my B.Tech in EEE. Fastai, being my very first course in Deep Learning, it greatly helped to understand Neural nets from scratch without having the “this is a magic” feel. Fastai’s user-friendly syntax and never failing defaults have made deep learning a lot simpler and efficient.
I use fastai to implement deep learning in different fields like, to predict chest cancer or diabetes, NLP in Tamil language and also learning to implement deep learning models in independent systems, for instance using Raspberry Pi. It would be great if the 2020 course covers deploying models as a stand-alone system.
Thank you @jeremy for the invitation to the 2020 course. It is great to be a part of Fastai’s forum.
Hi everyone,
I’m computer science student from Finland. I have followed what’s happening in fastai since the last course but because I started college and now joined a startup it’s sometimes hard to find time to talk here and read comments by other people. But as Jeremy hoped in the message I try to take the habit of reading and writing here before the course starts.
People have been really thankful for my notes from past courses and it always makes my day to hear nice feedback. Personally I’m never satisfied with the quality of the notes after writing them but I’m happy to hear that they help other people. The plan is to again write notes because that’s a great way for me to learn but I hope that other people also could write their own versions which are hopefully even better than mine.
I have seen Jeremy talking on Twitter about how to create own blog but I probably use something else. Medium is not even an option anymore because they have the paywall thing which drives everyone crazy. I’m planing to use Notion because I have moved most of my notes from last year there too and the plan is to keep it as public notebook. I like some of the features they have and that’s why I’m sticking with it instead of creating own blog.
Anyways I’m excited that the course will be starting soon.
Hey everyone, I joined about a year ago and worked through the 2019 versions of the course last year. Since then I’ve been working on Kaggle competitions and working with other members of the fastai community on various projects.
I have two bronze medals on Kaggle and would like to continue improving this year. If anyone would like to team up I’m open to that. Currently I’m working on the Google Q&A Challenge which wraps up in ~two weeks.
I stream my work on Twitch every weekday at 2pm EST and 9pm EST. Stop by if you want to chat about fastai, Kaggle or deep learning!
Hi everyone!
My name is Akash. (https://mobile.twitter.com/akashpalrecha98) I’m a third year Mathematics student at BITS Pilani, India. I’m also the AI Lead for a space tech startup called Pixxel .We build small satellites that will capture high resolution Hyperspectral Imagery of the earth, something that hasn’t been commercially available before. We were incubated at NASA/Caltech’s Jet Propulsion Lab last year for 3 months. Our first satellite launches in June 2020!. At this point, me and my team are just starting to teach ourselves GeoSpatial Deep Learning)
I started my coding / deep learning journey back in the summer of 2018 with Andrew Ng’s ML course on Coursera. In June 2018 I started doing FastAI’s Deep Learning course but was quickly bottlenecked by my lack of experience in coding / python / numpy etc. I took a sabbatical from the course for a while to level up my python skills, took CS231n and got back to doing FastAI in the middle of my 3rd semester in college.
I made a lot of mistakes while following the course in 2018, like not trying to reproduce the notebooks, trying to move very quickly through the course, etc.
In the 2019 version, I tried to correct all of that and made sure I deeply learn everything that Part 1 had to offer.
I also headed deep into Part 2 of the 2019 course (which I am still doing, at lesson 11-12). I am doing this part of the course at a deliberately slow pace. My objective not being to complete it, but to learn the essence of the process that Jeremy takes us through. Instead of reading just the research papers that Jeremy outlines during the course, I’ve built a habit of following all the latest AI research and reading new literature regularly. I’ve also tried working on 2 of my own research projects and the process of troubleshooting them, getting those approaches to work, etc would have never worked without having done FastAI.
For the most part though, I’ve been a very passive observer in most online forums that I take help from. I’ve contributed in some ways I believe to the FastAI forums, but I believe it’s not nearly enough. I have a habit of just chugging away through code on my laptop, figuring out everything myself solo, and then just forgetting about sharing that knowledge. I believe this habit is shared by a lot of other great developers too.
I like @Lankinen’s idea of using Notion. I have my personal website currently pointing to my page on Notion (akashpalrecha.me). I’m still figuring out what I’ll finally use to write more though. I’ve setup another website https://akashpalrecha.github.io (I found FastAI’s template on Jekyll’s website and happily adopted it ) meanwhile. The issue with Notion is that sometimes it’s too slow to load up and the urls aren’t really that descriptive too. This may make it harder for search engines to find it.
I plan to write more about achieving very specific things in DL, like, maybe how to very quickly build a Resnet with a Mish activation without going through a lot of trouble, etc. In writing these posts, I will try and highlight methods you can use in general at other places. So far the only blog I’ve written sits on Matplotlib’s official blog : https://matplotlib.org/matplotblog/posts/an-inquiry-into-matplotlib-figures/
This was a long post!
Anyways, my semester plans have to change wildly now after taking all of this into consideration.
Thanks a lot for the opportunity!
my name’s Oliver Müller. I studied applied mathematics. Programming was always a hobby of mine, I like the idea of implementing theoretical ideas and actually see them do something.
My start into machine learning was the book " Python Machine Learning - Second Edition" (Sebastian Raschka, Vahid Mirjalili). Python was already my go-to language, but this book introduced important libraries like numpy, matplotlib and the jupyter environment.
By chance I found the fastai course (2018) and was hooked. I remember the excitement of trying out those seemingly simple concepts to produce such astonishing results. Since then I followed fastai, completing both 2019 courses.
I’ve been told that I’m good at explaining complex subjects in an easy to understand way. As such I think I can contribute the best by writing summary articles about the lectures. Probably multiple posts per lesson, given their length and information density.
With that in mind: Is there any reason not to use medium? I know the paywall is frustrating. But as an author one can simply choose to not use their partner program and therefor not be listed behind the paywall, right?
My name is Ramesh Sampath. I work as ML Engineer and spend most of my time doing error analysis and collecting data where my model doesn’t do well. I come to FastAI for refugesince FastAI V1 days and very thankful for this community that Jeremy and Sylvian have built. Also had a chance to collaborate with few of my fellow FastAI learners in the past couple of years. Looking forward to reconnect with familiar faces and make new friends in FastAI Community.
I am currently based in Yangon, Myanmar. Currently I am working as an independent contractor for building ML/DL solutions. I and my colleagues started a deep learning startup at MIT delta V which provided training and testing visual data for evaluating visual perception systems of self-driving cars. Unfortunately, we had to shut it down when we ran out of funds. Before that, I was a Master’s student at WPI, doing research on using DL systems in education settings.
I love contributing my DL/ML knowledge back to community and I have been holding AI discussion groups here in Yangon to promote students and professionals to engage in technical discussions and feedback sessions related to topics in ML/DL.
I want to use this course as a platform for educating students here in Yangon. I want them to understand the source code of how fastAI is written (which I personally like) and encourage them to take things apart, debug and experiment them on their own to understand DL/ML systems via programmatic implementations (rather than inundated with a lot of math equations)
(which was not too long ago). Once it’s done I know I’ll be trying to include it in the study group
), hosted study groups locally and also on TWiMLAI and have also written blog posts on fast.ai forum explaining some of the concepts that i understood.
.
) meanwhile. The issue with Notion is that sometimes it’s too slow to load up and the urls aren’t really that descriptive too. This may make it harder for search engines to find it.