Find the most similar images in your dataset (Feature extraction, PCA, distance ML4A Notebook)

How do we find the most similar images in our dataset? Could we eventually use this to determine how diverse our dataset is? Let’s walk through a notebook taken from Gene Kogan’s ML4A Guides: ml4a.github.io...
Slightly modified version of the ML4A notebook is here: github.com/dvs...
Join this channel to get access to perks:
/ @artificialimages
Learn more about machine learning for image makers by signing up at mailchi.mp/da9...
Join the Artificial Images Slack channel: join.slack.com...
artificial-imag...
/ bustbright
bustbright.com/
/ dvsmethid
/ dvsch
dvschultz.gith...

[@shivamgoel0897]

Super! This is exactly what I was looking for

[@bigdatavisualization7970]

Great job!!! Helped me a lot. Thank you very much

[@vikramsandu6054]

Its amazing. thanks for the video. I really needed this one.

[@dikshabagade5542]

Thank you so much dude!

[@rob26r]

How does it handle similar images that are angled (150, 180, 210 degrees)?

[@karthikr5884]

All the angled images will have similar feature vectors from the VGG pre-trained model.

[@PinjuSaikiaGuwahati1]

how we can use our own image folders instead of this dataset?

[@smokinep]

thanks .. the video is pretty useful for reverse image search. Is it feasible to training a daily model of new images and append it to the pickle outputs ?

[@karthikr5884]

I would like to correct you on the model training that no model is trained here. When you include new images, you must recompute the PCA for the entire images and finally save the pickle outputs.

[@projectgroot9338]

how to use the pickle file in another notebook

[@ammarhaiderjafri9390]

Hi, Github link is not working, any working link can you share or code here ?

[@PadminiMansingh]

May I have ur mail-id sir