General course chat

Aamir · January 2, 2020, 9:30am

I am so pumped up right now after listening to first lecture. I have couple projects in my mind. Wish me good luck to make it happen.
love u

Aamir · January 2, 2020, 9:31am

be consistent man. sure you will do

mrfabulous1 · January 3, 2020, 8:46am

Hi pratikw hope all is well!
I am sure that if you follow Jeremie’s advice and type in everything by hand, experiment and get to lesson 8 you will come out better.

Cheers mrfabulous1

Tammy · January 5, 2020, 6:21am

Hi ,

Is there any difference in topics covered in v2 and v3. I initially took Andrew Ng Deep Learning course now looking to take the course in Fast ai and confused between v2 and v3. Digging deep I found topics like Darknet covered in v2 which I am more interested in, I am planning to do a project which requires YOLO algorithm for realime detection.

Everyone seems to advice to take Practical Deep Learning for Coders 2019 first.
Does v3 covers topics related to realtime object detection??

ishitizkhanna · January 16, 2020, 3:49pm

I couldn’t find the option to create topic so,i am publishing my doubt here,folks

While implementing a Neural Network on a tabular data,i m facing this error in learn.predict function

KeyError Traceback (most recent call last)
/usr/local/lib/python3.6/dist-packages/pandas/core/indexes/base.py in get_loc(self, key, method, tolerance)
2896 try:
-> 2897 return self._engine.get_loc(key)
2898 except KeyError:

pandas/_libs/index.pyx in pandas._libs.index.IndexEngine.get_loc()

pandas/_libs/hashtable_class_helper.pxi in pandas._libs.hashtable.PyObjectHashTable.get_item()

KeyError: ‘Review’

During handling of the above exception, another exception occurred:

KeyError Traceback (most recent call last)
10 frames
/usr/local/lib/python3.6/dist-packages/pandas/core/indexes/base.py in get_loc(self, key, method, tolerance)
2897 return self._engine.get_loc(key)
2898 except KeyError:
-> 2899 return self._engine.get_loc(self._maybe_cast_indexer(key))
2900 indexer = self.get_indexer([key], method=method, tolerance=tolerance)
2901 if indexer.ndim > 1 or indexer.size > 1:

pandas/_libs/index.pyx in pandas._libs.index.IndexEngine.get_loc()

pandas/_libs/hashtable_class_helper.pxi in pandas._libs.hashtable.PyObjectHashTable.get_item()

KeyError: ‘Review’

Can somebody help me out?

am.sharan · January 21, 2020, 4:00am

what is callback? why are there so many different callbacks ? how to use them?

mrfabulous1 · January 21, 2020, 10:34pm

Hi am.sharan hope all is well!

The callback system in the fastai library allows you to execute any code at specific times in the training. This can be extremely useful if you want to
record some parameters (for instance metrics, gradients)
use a condition to change a parameter (for instance change the learning rate if the validation loss hasn't improved)
make a schedule of parameters (for instance changing the probabilities of the dropout layers as the training goes)

The above quote is from a notebook in this link Using the callback system in fastai

I also found these posts useful to. https://medium.com/@edwardeasling/implementing-callbacks-in-fast-ai-1c23de25b6eb
https://medium.com/ai³-theory-practice-business/callback-fastai-09-436615d07186

Cheers mrfabulous1

Miske · February 6, 2020, 12:11pm

Hello everyone. I competed in my first ever Kaggle competition (MNIST), and now i’m looking for a similar Kaggle beginner friendly competition in image classification. any suggestions ?

piby4 · February 6, 2020, 1:20pm

@Miske
If you want to try out another MNIST kind of dataset try-
MNIST like datatset for Kannada handwritten digits ( https://www.kaggle.com/c/Kannada-MNIST )

Else you want to graduate to more ‘real-image’ kind of competitions, here is a list from the archives ( this is a little dated - there are newer ones though )

About	Link
State Farm Distracted Driver Detection ( Can computer vision spot distracted drivers?)	https://www.kaggle.com/c/state-farm-distracted-driver-detection
Airbus Ship Detection Challenge - Find ships on satellite images as quickly as possible	https://www.kaggle.com/c/airbus-ship-detection
Cdiscount’s Image Classification Challenge (Categorize e-commerce photos)	https://www.kaggle.com/c/cdiscount-image-classification-challenge
Aerial Cactus Identification (Determine whether an image contains a columnar cactus)	https://www.kaggle.com/c/aerial-cactus-identification
Dog Breed Identification (Determine the breed of a dog in an image)	https://www.kaggle.com/c/dog-breed-identification
iMaterialist Challenge (Furniture) at FGVC5(Image Classification of Furniture & Home Goods.)	https://www.kaggle.com/c/imaterialist-challenge-furniture-2018
Plant Seedlings Classification(Determine the species of a seedling from an image)	https://www.kaggle.com/c/plant-seedlings-classification

Miske · February 6, 2020, 4:47pm

Thank you very much . This reply helped.

vaio05 · February 8, 2020, 4:37pm

Hi guys,
I am not sure if this is the right place to ask, I will give it a shot.

This course talked a lot about “progressive resizing” the input image size, when solving a image classification task. I want to implement this in my thesis I am writing, Jeremey once mentioned a paper regarding this topic, but I can’t find any literature about this method. Could you help me out?

Thanks!

pinaki · February 10, 2020, 10:57pm

Hi,
Is there an issue with the files.fast.ai servers ?
path = untar_data(URLs.MNIST_SAMPLE); is failing with a connection timeout error – while the ones hitting s3 work just fine.
http://files.fast.ai/data/examples/mnist_sample – doesn’t work.
https://s3.amazonaws.com/fast-ai-imageclas/mnist_png – works
ConnectTimeout: HTTPConnectionPool(host=‘files.fast.ai’, port=80): Max retries exceeded with url: /data/examples/mnist_sample.tgz (Caused by ConnectTimeoutError(<urllib3.connection.HTTPConnection object at 0x7f67705483c8>, ‘Connection to files.fast.ai timed out. (connect timeout=4)’))

pinaki · February 10, 2020, 11:04pm

hi guys seems like your SSL certs expired ?
the file still exists – just the ssl certs are bad ?

lwang94 · February 12, 2020, 12:30am

hey so I’m getting through the fastai deep learning courses and I was wondering if there was any one that talked about choosing the right metrics for the types of data you may encounter?

Miske · February 13, 2020, 10:09am

Hello everyone , i am looking for a mentor .In my social surroundings there is no who can be my mentor. If someone can be my mentor please give me a reply.I believe i can bring a lot of value to someone who is my mentor.

vdemario · February 13, 2020, 9:37pm

Hello everyone, does anyone know if there has been an effort to translate the content of the course to Brazilian Portuguese? I’ve searched through the forums but didn’t find something like that.

Related: I’d like to create a presentation covering the beginning of the course to submit to a conference (in Portuguese) but I don’t know if this would be ok or not. What’s the best way to ask for permission to adapt the material like that? I’m not sure which license covers it (this one? https://github.com/fastai/course-v3/blob/master/LICENSE) and I’d like to do it only if it’s considered ok by the authors.

lubos · February 19, 2020, 8:18pm

Hello everyone!

I have a couple of beginner questions.

I have a csv file containing: 1st column: file names of images; 2nd column values 0 - 100. I want to create a DataBunch to do an image regression. Is the following code correct for creating a suitable DataBunch?
data = (ImageList.from_csv(path, ‘files_and_values.csv’)
.split_by_rand_pct(valid_pct=0.2, seed = 42)
.label_from_df(label_cls = FloatList)
.transform(tfms=[], size=224)
.databunch())
data = data.normalize(imagenet_stats);
How to create a DataBunch for doing image regression with the csv file containing additional 3rd and 4th column of numbers.
Is it possible to use y_range = [50,102] with image regression, likewise performed in learn = collab_learner(data, n_factors=50, y_range=y_range)?

Thank you!

Lubos

lchunleo · February 22, 2020, 2:58pm

Can i check for fast.ai is there a certificate course?

zahra4fact · February 22, 2020, 3:38pm

hi, thanks for your course in advance .
actually I have a problem with this line …
newpath=get_file(‘vgg16.h5’,file_patch+‘vgg16.h5’,cache_subdir=‘models’)
model.load_weights(newpath,reshape=True)

when I run this I get the error for unmatching wight shape and size:
" Weights must be of equal size to apply a reshape operation. Layer conv2d_1’s weights have shape (3, 3, 224, 64) and size 129024. The weights for loading have shape (64, 3, 3, 3) and size 1728. "

How can I handle it ? anyone knows ??

banksiaboy · February 24, 2020, 12:32am

Andrew Glassner’s “Deep Learning: A Crash Course” from Siggraph is very good. Just above the algebra.

Computer Graphics nerds might remember Andrew from the ‘Graphic Gems’ books.
He also has an e-book(s) on the subject - https://github.com/blueberrymusic/DeepLearningBookCode-Volume1 has appropriate links.