MNIST Database

how to get a qt redneck gf?

This is basically 'Hello World'. Go for it, but don't have unrealistic expectations for what it means for your portfolio.

I did it in javascript

Star with your sister.

MNIST is baby tier. It's trivial to do much more complicated classification tasks using pre trained models/transfer learning.

If you're interested in digit recognition and want to do something cooler, you could train a model to actually read handwritten text. Pretty easy to do either with an LSTM based model or a fully convolutional model. Look up Connectionist Temporal Classification for how you'd do the reading part. You could also pretty easily make a Flask app and host it on AWS. People upload images of handwritten text and the model transcribes it. That's a way more interesting project than MNIST classification.

If you're interested in classification stuff, find a more exciting dataset. If you're limited on compute, you should look into the Oxford pets dataset or Food-101. Good smaller datasets.

It's also not much of a stretch to go from CNN classification to object detection.

>Any recommendations for how to get started?
If you're unsure of where to start in terms of learning, you should look into Andrew Ng's deep learning course on coursera (Tensorflow) or fast.ai (pytorch).

If you just want to jump into classification, I'm pretty sure that's lesson 1 of fast.ai.

neuralnetworksanddeeplearning.com/chap1.html

there are literally 1 hours lecture on fast.ai to get 99% accuracy on MNIST
any brainlet can do this within

maybe you niggers can relax with the "baby tier" faggotry? op wants to do a project, why the fuck you gotta act so smug

f*et...

Hot.