If any international fellow doesn’t have a group (i.e. you didn’t update this spreadsheet, please use this thread to say so, so that other fellows without a group can team up with you. Currently group 163 has only 3 people, so you could join that group.
I could join them and strengthen the groups numbers if its alright my current group has 6 members
Now streaming. Let me know if any issues. You can see your class mates 
1 Like
No issues here
Working here
This piece of code stop working after the git pull
arch=resnet34
data = ImageClassifierData.from_paths(PATH, tfms=tfms_from_model(arch, sz))
learn = ConvLearner.pretrained(arch, data, precompute=True)
learn.fit(0.01, 3)
try to clear tmp directory
@jeremy The spreadsheet linked is not accessible. It is asking for permission.
3 Likes
Oops sorry! Fixed now.
BTW, please use the weekly wiki/discussion thread for in-lesson chat, not the youtube chat, so that everyone can benefit from your discussion. 
1 Like
The live stream will be at https://www.youtube.com/watch?v=nhEQrzf1pIQ . Please use the lesson forum thread for discussion - I’ve disabled the live chat on youtube, since I don’t want other students to miss out on your valuable questions and comments!
Sorry corrected URL https://www.youtube.com/watch?v=_VpaKaMyjqI
Hey guys! Why the size is fixed in 224? bs means what?
Batch size, size of a chunk which goes to GPU for computation
1 Like
Size is not fixed, you can also use 299. Only think is that size of all images should be consistent
1 Like
So all images will be resized to 224x224? Thank you for the answers!
How do we decide what size to start with or use consistently? Or is that experimental based on our observations of loss/accuracy?
My understanding is that CNNs do not depend on image size, filter can be applied to any image sizes but more the size of image more is the need of compute and memory requirement.
Yes, happens on the fly
Anyone from the international thread in Kenya or Uganda, I will be visiting both this week. If in Uganda I will be visiting Kampala and Nairobi in Kenya. If there is anyone it would be great to meet up.
1 Like
FYI it’s better to use the main wiki/discussion thread during class, so that all can benefit.