https://github.com/pvlbzn/kot-or-not

Logistic regression (shallow neural network) which predicts whether input image is a cat or not.
https://github.com/pvlbzn/kot-or-not

classification-algorithm image-classification logistic-regression machine-learning neural-network numpy

Last synced: 8 days ago
JSON representation

Logistic regression (shallow neural network) which predicts whether input image is a cat or not.

• Host: GitHub
• URL: https://github.com/pvlbzn/kot-or-not
• Owner: pvlbzn
• Created: 2017-09-13T18:53:45.000Z (about 7 years ago)
• Default Branch: master
• Last Pushed: 2017-09-13T19:28:44.000Z (about 7 years ago)
• Last Synced: 2024-10-30T03:32:18.550Z (22 days ago)
• Topics: classification-algorithm, image-classification, logistic-regression, machine-learning, neural-network, numpy
• Language: Python
• Homepage:
• Size: 2.46 MB
• Stars: 1
• Watchers: 2
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

        
# Kot or Not

Kot or Not (from Russian word Кот, a Cat, wordplay) project is an implementation

of a [logistic regression](https://en.wikipedia.org/wiki/Logistic_regression)

algorithm using `numpy` for matrix arithmetics.

Algorithm uses [`sigmoid`](https://en.wikipedia.org/wiki/Sigmoid_function)

function, forward and backward propagations as well as vectorization techniques.

## Data

Current implementation has training set with cardinality 209. Images scaled

down to `64 x 64 x 3` and processed as a feature vector `(64 * 64 * 3, 1)`.

## Usage

```

usage: main.py [-h] [-d] [-r] [-i INPUT]

Cat or not neural network

optional arguments:

  -h, --help            show this help message and exit

  -d, --dump            use model dump

  -r, --retrain         retrain model

  -i INPUT, --input INPUT

                        path to input image

```

To classify your image first you need to train a model

```

python3 main.py -r

```

The model will be dumped into a flat text files: `weights` column vector

and `bias` literal.

After training `--dump` can be used together with `--input` image path

```

python3 main.py -d -i images/moon.jpg

```

Baby can Moon is certainly a cat, and `kot-or-not` NN things the same.

## Roadmap

* Implement a self learning feature from input images