What could be improved? Machine Learning Project Retrospective | Airbnb Amenity Detection Part 9

At the end of any project, it's essential to look back and ask two questions: What went well? and What could be improved? In this video, I ask these questions looking back on my machine learning project to replicate Airbnb's amenity detection.
If you're not sure what's going on, make sure to check out the rest of the videos in the series.
Now this project is over, it's time to move onto the next one!
Links:
My machine learning course - dbourke.link/mlcourse
Blog post version of the project - www.mrdbourke.com/airbnb-amen...
Example code on Google Colab - dbourke.link/airbnbcode
Code on GitHub - github.com/mrdbourke/airbnb-a...
All of my notes/ideas in Notion - dbourke.link/airbnb42days
Airbnb Amenity Detection article - / 05qperz3a4
Check it out for yourself - airbnb-amenity-detection.apps... (note: may be taken down due to hosting costs)
All videos in this series:
Full playlist - dbourke.link/airbnbplaylist
Part 1 (overview) - • Using Machine Learning...
Part 2 (data exploration) - • Using Machine Learning...
Part 3 (data manipulation) - • Replicating Airbnb's M...
Part 4 (building a small dog model) - • I got Detectron2 worki...
Part 5 (experiment tracking) - • Every model I build ha...
Part 6 (modelling experimentation) - • 34,835 training images...
Part 7 (building a big dog model) - • Results from 18-hours ...
Part 8 (getting my modelling deployed) - • I got my machine learn...
Part 9 (project wrap up and retrospective) - • What could be improved...
Note: I am not affiliated with Airbnb nor are they involved with this series of videos in any way.
Connect elsewhere:
Email updates on my work - dbourke.link/newsletter
Web - dbourke.link/web
Quora - dbourke.link/quora
Medium - dbourke.link/medium
Twitter - dbourke.link/twitter
LinkedIn - dbourke.link/linkedin
#machinelearning #datascience #deeplearning

[@arjit1117]

I am a simple man, I see Daniel's video, I click.

[@shawnjames3242]

Same

[@mrdbourke]

I am a simple man too, I see this comment, I click

[@negar5994]

I liked the analogy between replicating the article and following a recipe and tweaking it a bit. It made me more inspired to work on my own project.

[@sebah216]

Just 1 word, AWESOME. Congrats Daniel.

[@camerontillett8295]

I'm doing your Machining learning course, the way you teach is amazing. The only course I've come across that actually makes me think while I'm learning.

[@mrdbourke]

So great to hear Cameron! Thank you for the kind words my friend

[@arunpandian7]

Its just since yesterday, you quickly become my inspiration. Thank you so much for your tips on productivity and self learning. I hope I will see my productivity boosted up in near future. I have to say, you are great content creator and you deserve a lot.

[@mrdbourke]

Thank you Arun, you're very kind. And that's so good to hear my friend. Keep getting after it.

[@chaskastuff8939]

You are one of best machine learning instructor. Thanks for such nice videos. I am currently doing your machine learning course. It's amazing. 👍👍👍👍👍

[@mrdbourke]

woah, thank you so much Chaska, that means a lot. glad you're enjoying the course

[@zzheng4401]

amazing job!!!

[@fahadreda3060]

Great video Daniel,Thanks, what type of Keyboard you are using ?

[@amircodes]

I do really love your channel content.

[@mrdbourke]

Thank you code party, I appreciate that!

[@harikiranm5756]

such an inspiring cool guy

[@adybrain7693]

Your Mr Motivator for me, thanks! Quick Q, how much web dev skills did you need create the site? Any recommendation or did you adhoc skills along the way? Thanks.

[@linusjohansson3164]

I am about 1000 hours into my journey to master machine learning starting from scratch. I have completed some courses online, read books and watched hundreds of videos on youtube. These videos are some of the very best imo! Suggestions: Another project like this or smth where you go more in depth about transfer learning, hyperparameter tuning for DNNs , error analysis or advanced data cleaning/preprocessing.

[@mrdbourke]

Hey Linus! Thank you for the kind words my friend. And I appreciate the suggestions for the future!

[@shawnjames3242]

Hey Daniel, this was an amazing series of videos! Can you make more videos like the one where you showed how you worked on the titanic dataset ?

[@mrdbourke]

Thank you Shawn! Glad you enjoyed it. Great idea about the Titanic dataset video. You mean more Kaggle-style videos? Or just any dataset and how I’d approach it?

[@kailkuhn]

Hey Daniel, what app did you use to create the graph you're showing in 11:30 (the one you have on your blog too)? Btw, great series dude, seeing this project thru to deployment and being able to organize yourself to do so. Good job!

[@mrdbourke]

Hey Kail, I created the graph using Whimsical: whimsical.com/ And thank you for the kind words my friend, I really appreciate it.

[@amrindersingh1292]

I have two goals to achieve in my life. Seeing Daniel's videos motivates and encourages me to work intensively hard for both of them.

[@mrdbourke]

Get after it Amrinder. I'm cheering for you!

[@amrindersingh1292]

@@mrdbourke yes! I am on it.. Thanks bro!

[@shawnjames3242]

@@amrindersingh1292 is it fitness and machine learning

[@amrindersingh1292]

@@shawnjames3242 u got it right! It's running and machine learning.

[@kingberz]

Why is GPU required here in first place? This could be such that -> load image to trained model -> spit out detection. Predicting or classifying on a trained model does not require GPU? They are only needed when training? Did I get something wrong?

[@wouldntyaliktono]

This feels a lot like a long-form job application

[@orrimoch5226]

Geart stuff! it's always the inference time that screw you :-)

[@jeffrey5602]

have you thought about building ur own ml workstation?

[@mrdbourke]

I have yes, been looking into just this week. Might be worth it in the long-run. This article is great: medium.com/the-mission/why-building-your-own-deep-learning-computer-is-10x-cheaper-than-aws-b1c91b55ce8c

[@jeffrey5602]

@@mrdbourke although colab for example is potentially a cheap platform to development it cant match the responsiveness of a local jupyter instance. Plus you can run longer training sessions for a few days straight. Just take my advice on avoiding any kind of rgb if the workstation is in the same room as your bed :D. Iam currently running 2x 1080ti's 64gb ram and 16 core amd threadripper. I was only forced to use AWS for some of the udacity nano degrees

[@SerJG]

Is quarantine over? There is no new videos with exercises at home, I really get used to them))

[@shawnjames3242]

That was for 30 days i think

[@mrdbourke]

Quarantine isn't over but Shawn is correct, it finished off as a month of movement.

[@SerJG]

Daniel Bourke I see :) thank you for that one month movement videos

[@PavanMuppala]

Bro, I have difficulties using git code/scripts, there are so many files and I don't understand how they are arranged and how to get them running, I have searched for tutorials but I didn't find any, could you point me to any resources for that.

[@mrdbourke]

Which files are you trying to get running? Have you been through GitHub's tutorials? lab.github.com/

[@manoharlakshmana6171]

Hi Sir, can you make a short video where you explain how your code is stored in GitHub ? So that we can replicate it ? Please it would be helpful for people who want to start their career and are beginners in the field. Thank you. Hoping to get a reply from you soon.

[@SaiNikhileshReddy]

Bro which is ur task manager I know trello but you use another software to take course notes and tasks

[@mrdbourke]

Notion

[@SaiNikhileshReddy]

@@mrdbourke thanx bro

[@denismerigold486]

Can I get a job in machine learning without a degree?

[@prepboost6953]

Is that surface pro .....?? You use iPad pro or surface pro for light weight tasks

[@mrdbourke]

I use: MacBook Pro for most computing + iPad Pro for reading/writing

[@apkingboy]

Why not use YOLO bruh?

[@shawnjames3242]

Nice beard! 🧔

[@mrdbourke]

Viking mode: engage.

[@thedeegan]

Well... in all honesty... it isn't your model... it's facebook's model... still awesome project though