264 - Image outlier detection using alibi-detect

Code generated in the video can be downloaded from here:
github.com/bns...
Outlier detection using alibi-detect
Alibi Detect is an open source Python library focused on outlier, adversarial and drift detection. The package aims to cover both online and offline detectors for tabular data, text, images and time series. The outlier detection methods should allow the user to identify global, contextual, and collective outliers.
pip install alibi-detect
github.com/Sel...
Documentation: docs.seldon.io...
We will be using VAE based outlier detection. Based on this paper:
arxiv.org/pdf/...
The Variational Auto-Encoder (VAE) outlier detector is first trained on a batch of unlabeled, but normal (inlier) data. Unsupervised training is desirable since labeled data is often scarce. The VAE detector tries to reconstruct the input it receives. If the input data cannot be reconstructed well, the reconstruction error is high and the data can be flagged as an outlier. The reconstruction error is either measured as the mean squared error (MSE) between the input and the reconstructed instance or as the probability that both the input and the reconstructed instance are generated by the same process.
Data set info: openaccess.the...
Data set link: www.mvtec.com/...

[@brendensong8000]

Another outstanding video! Thank you for making and sharing this video and the codes!! Your lectures and explanations are by far the best out of all the videos I've seen!!!

[@jacobusstrydom7017]

Very cool, I haven't watched all of your recent videos so I need to do some catching up. Great job as always

[@iyad863]

Justice has been served

[@irfannoordin2245]

Hello, does anyone know how to use a different input image size for this model . I keep getting InvalidArgumentError: required broadcastable shapes [Op:Sub]

[@kenanmorani9204]

Thank you for an amazing video and explanation.

[@romanprochazka179]

Hello, thank You for sharing. Really nice content!

[@sunmhaske4157]

Nice video sir . Thanks you so much ...

[@suryaravi3907]

Hi Sir, Thank you for all amazing way of teaching. This video is very useful since I am researching in Alibi-detect lib. I went through seldon videos on Alibi which was not as explanatory as yours with live coding.
One request Sir, can you make video for drift detection using alibi detect library, with other dataset not cifar10 as mention in Alibi doc. It will be very useful for me.

[@davidvexler7660]

Hello, thank you so much for the video! It has been the best help on this topic I have seen yet. My question is, however, how do you open xz files(The carpet) on a window?

[@shanrockz4104]

why loss value in negative , ex.20090.8508 ? Is it good sign?

[@kumarsyamala]

Hi, I have downloaded in the dataset, we have only good images in the training folder, how did you get the Bad images!!

[@tiwalademodupeusman3452]

What is the best outlier detector method for monitoring disease progression using longitudinal medical images? For example on each a healthy image there are no biomarkers but as the disease progresses more different biomarkers appear. There could be up to 5 new and different biomarkers on an image.

[@DigitalSreeni]

Is that a outlier problem or a segmentation problem?

[@assafkeller]

Hi, did you solve the Saveing error? i tried with python 3.7 same problem

[@tonya4092]

Thanks!

[@DigitalSreeni]

Thank you very much Tony.

[@tonya4092]

@@DigitalSreeni No issue, you deserve it. Any pointers/examples using tensorflow's ImageDataGenerator with alibi-detect's OutlierVAE?

[@Guide4Ever]

Why is there 3 instead of 32 as 1st parameter of Conv2DTranspose? 13:35

[@DigitalSreeni]

It is the output layer and I want my output to have 3 channels giving a 64x64x3 array (image).

[@kingping8386]

Can we have a tutorial about text-to-image gan soon?

[@abd_alim]

Hello sir, can we use this for outlier detection in videos? What is the inferencing speed?

[@DigitalSreeni]

You can use this for videos, not just single images. I cannot comment on speed as I haven't tested and also it depends on the image size.

[@dr.aravindacvnmamit3770]

"IndexError: index 9 is out of bounds for axis 0 with size 0" I am getting this error, Since I used Polyp image datasets ? Can anyone helpme out

[@dr.aravindacvnmamit3770]

I have solved it 🔥🔥🔥. Thank you Srini Sir. The line 156&157 which one you have commented even that will work I did 🎉🎉🎉

[@NMCSDAKSHADINESHSHENOY]

Congratulations sir

[@areezabhanbhro6005]

How did you solved this error ?

[@humanity5175]

Hello everyone ☺️,
I have a question :
Which is the best ML platform used for industrial purposes ?

[@DigitalSreeni]

I'd be surprised if anyone can give an answer to this question. 'Industrial purposes' is a broad term as it depends on the application. If you are looking for pathology applications you may find someone specialized in that domain, similarly for other applications.

[@sofienabidi6112]

thank you for all this amazing tutorial, after trying it, i found that the saved model (from alibi_detect.utils.saving import save_detector, load_detector works) is too heavy for edge devises (550MB). is it possible to use also alibi detect to extract the coordinate of anomaly in order to localize it ?

[@thepurplecow3757]

Where is the bad images coming from? The training directory that I downloaded from the site contains only the good images!
Also, when running the first part, it spits out "AttributeError: module 'tensorflow.keras.optimizers' has no attribute 'legacy' " message. 😞

[@thepurplecow3757]

Oh, the message of "AttributeError: module ~~~" is solved by upgrading my tensorflow.

[@Guide4Ever]

Current path for tensorflow.keras.layers has been moved to tensorflow.python.keras.layers

[@jhonmunoz3535]

I get an error when I want to train the model, it doesn't let me move forward, it happened to you