HI @harish3110 great Idea, you can count on me.
1 Like
@harish3110, I’d also like to collaborate
2 Likes
Why not? I’m also interested
2 Likes
@Albertotono, @ganesh.bhat, @Blowoffvalve, @kofi Let’s get on a zoom call sometime soon as per everyone’s availability and get this going then! I was thinking we can do an end to end project on an image classification dataset.
I was looking at the APTOS 2019 Blindness Detection in particular. It looks like a decently tough dataset that was completed just 8 months ago. It’s a diabetic retinopathy classification project with retina images of 4 different gradations. It’s evaluated on the quadratic weighted kappa metric. Just another metric that is available on scikit-learn directly.
The top performer’s results of this competition and many more have been well documented in this blog.
It’s a fairly large dataset, but manageable for our experiments and learning. If you guys find any other interesting image datasets we can discuss that during the call and go ahead! 
What time?
I can do same time as the class I.e. 7 PM SFO time or 8:30 AM SFO time
1 Like
That works for me as well!
1 Like
@ganesh.bhat, @harish3110, what date are you looking at?
1 Like
In case we have a Fastbook meeting as scheduled at 7.30 am IST on Sunday then maybe after that? If not we could have it at the same time slot.
1 Like
If we can get >5 likes to this reply, I’ll host it 
5 Likes
Yes, agree to it.
@harish3110 Are we scheduled to meet in in an hour
@Blowoffvalve Didn’t think people were available due to the lack of confirmations and hence didn’t set it up. But sure, we can have one if you’re interested. Anyone else interested can direct message me!
@harish3110 how did you download the data from kaggle in colab?
!kaggle competitions list -s aptos2019-blindness-detection
The above command is downloading all files without the folder structure of train and test.
!mkdir folder_name && cd folder_name
!kaggle competitions download -c aptos2019-blindness-detection
That should do it after you’ve got your Kaggle api installed and your credentials added to ~/.kaggle.
Try %cd '/content/aptos' and then try downloading!
I had to do os.chdir('aptos) which did change the directory but it is still downloading png & zip files.
Not sure whats wrong.
Not sure then. It looks like it’s downloading the zip file and unzipping it as well. It seems to be a Colab issue, I use GCP.
Ok.
I got it setup on kaggle notebook. Much easier and no hassles.
I knew I’m late but did you try to update kaggle api with the github version and then download the dataset? In the past I too got this weird behaviour and it got fixed by updating kaggle api in colab.