Generating Images with Python and PIL for Machine Learning

preview_player
Показать описание
Sometimes it can be helpful to generate a large volume of JEPGs or PNGs to create synthetic data for machine learning training. In this video, I demonstrate how I used Python and PIL to create an image recognition dataset for one of my classes.

Рекомендации по теме
Комментарии
Автор

People new to this should use the module named Pillow, which is an actively maintained fork of PIL. You use it exactly in the same way as PIL, with "import PIL".

plagiats
Автор

Hello Jeff, I had a question about what neurons learn in each layer, for example in lstm networks in a time series. How can you know what they learn in the first layer or in the hidden layers?

benjamingalas
Автор

Thank you soo much, this is what I asked 🙏🙏

Creek
Автор

Hi, Jeff. I finished a Coursera course in Data Science. I would like to practice, where should I start with your videos? Thanks!

tatianaguanangui
Автор

How can I generate multiple PNG sequences in 32px x 32px from a collection of assets from a folder. Each asset should be a specefic space as per asset ( Tshirt + Hat + Glasses ) for example

gokulhbk
Автор

Hello professor.
Can I use binary cross entropy loss with one hot encoding in keras. I try I but it gave me good accuracy compare to categorical cross entropy. But I am confused I wonder if I am properly using the loss.

bediosoro
Автор

Hi sir, I have a question unrelated to this video. As we are progressing towards quantum computing, I wanted to know how we will see combo of machine learning and quantum computing. Would love to hear about some use cases from you. Thank you.

lakshay
Автор

Hello professor love your channel. A quick question. I have keras/tensorflow faster rcnn model I built up and I only saved the weights with save_weights during training. I'm trying to convert the model to tensorflow lite but the converter isn't working because it expects the model in the file but I only have the weights. Is there a way to still convert to lite using these weights. Like can I load the model like I did during testing and somehow save it from there and use it to convert to lite? Thanks and stay safe.

revantreddy
Автор

Hi Professor. I have a question about the docker image you create for funning your classes. How do I update that with new R/Python packages?

francycharuto
Автор

Great video professor!
How can I generate random NEW images based on a model trained by sample images. I want to do it for NFTs generation. What would you reccomend? TIA

PepeTostado
Автор

Can I take my family album and make a pil library to generate funny images of my family?

bluelantern
Автор

me tryna figure out why he started at 25001:

: (

jackjenny
join shbcf.ru