Computer vision and machine learning notebooks

Some algorithms I have implemented recently. Most of the vision notebooks are based off ideas from the book "Computer Vision - Algorithms and Applications" by Springer. The machine learning models are a bit of everything. Some basic, some more advanced stuff. You need to show a spectrum of knowledge.

Locally adaptive histogram

Qlearning

I wrote a basic game to test q-learning.

Anti-aliasing

Gan

based off https://arxiv.org/abs/1610.09585

Dropout

based off http://jmlr.org/papers/volume15/srivastava14a/srivastava14a.pdf model without dropout

model with dropout

Feature checker

Based off the idea from https://github.com/evilsocket/ergo (relevance.py). What feature is contributing what to the models results? If you are going to use this in production, set some random data points to zero during trainings as well.

Filters fixing noise

Gradient descent

Finding a local minima

A*

Search algorithm.

K-means clustering

LeCun CNN

Implemented LeChun CNN model, based off http://yann.lecun.com/exdb/publis/pdf/lecun-01a.pdf

Markov Random Field

Hough Transform

Basic implementation of hough Transform.

Auto encoder

Gan in numpy

Segmentation graph

MNIST generalization test

got the idea from https://arxiv.org/pdf/1611.03530.pdf , how much noise can a simple model handle and still do good evaluation on a noise free dataset?

the accuracy over time is based on the traing data.

Generalization

https://arxiv.org/pdf/1611.03530.pdf

Poisson image editing

Mostly a fork off this implementation, I fixed support for python3 and made it work with all the channels. Removed opencv depency as well.

Feature detection

Harris corner detector to find special features in each image. Using MSE to connect the special features.

Counterfactual regret minimization

Python implementation of the rock, paper, scissor section of http://modelai.gettysburg.edu/2013/cfr/cfr.pdf

Lucas–Kanade method, optical flow

Segnet

Had memoryerrors so this model was trained on only a subset of the training set (please hire me so I can build a computer for machine learning). I used the CamVid dataset.

Transfer learning

https://en.wikipedia.org/wiki/Transfer_learning

Autoencoder fixing image noise

Numpy rnn

Loosely based on iamtrask post. Maybe I wouldn't have coded this if he had used a linked list.

Name		Name	Last commit message	Last commit date
Latest commit History 4 Commits
README		README
A_star.ipynb		A_star.ipynb
Hough Transform.ipynb		Hough Transform.ipynb
LeCun_CNN.ipynb		LeCun_CNN.ipynb
MRF.ipynb		MRF.ipynb
Poisson Image Editing.ipynb		Poisson Image Editing.ipynb
README.md		README.md
anti-aliasing.ipynb		anti-aliasing.ipynb
auto encoder.ipynb		auto encoder.ipynb
autoencoder_noiseipynb		autoencoder_noiseipynb
check_model.ipynb		check_model.ipynb
counterfactual_regret_minimization.ipynb		counterfactual_regret_minimization.ipynb
dropout.ipynb		dropout.ipynb
feature_detection.ipynb		feature_detection.ipynb
filters_fix_image_noise.ipynb		filters_fix_image_noise.ipynb
gan.ipynb		gan.ipynb
generalization.ipynb		generalization.ipynb
generalization_test.ipynb		generalization_test.ipynb
gradient_descent.ipynb		gradient_descent.ipynb
k-means clustering.ipynb		k-means clustering.ipynb
locally adaptive histogram.ipynb		locally adaptive histogram.ipynb
np_gan.ipynb		np_gan.ipynb
optical_flow_lucas_kanade.ipynb		optical_flow_lucas_kanade.ipynb
qlearning.py		qlearning.py
rnn_numpy.ipynb		rnn_numpy.ipynb
segmentation_graph.ipynb		segmentation_graph.ipynb
segnet.ipynb		segnet.ipynb
transfer_learning.ipynb		transfer_learning.ipynb

2xic-archive/Notebooks

Folders and files

Latest commit

History

Repository files navigation

Computer vision and machine learning notebooks

About

Resources

Stars

Watchers

Forks

Languages