https://github.com/elucidation/tensorflow-cat-dog-classifier

Tutorial on using Tensorflow to train a cat/dog classifier, as well as building your own dataset to train on
https://github.com/elucidation/tensorflow-cat-dog-classifier

Last synced: about 1 year ago
JSON representation

Tutorial on using Tensorflow to train a cat/dog classifier, as well as building your own dataset to train on

• Host: GitHub
• URL: https://github.com/elucidation/tensorflow-cat-dog-classifier
• Owner: Elucidation
• License: mit
• Created: 2017-01-30T01:00:48.000Z (over 9 years ago)
• Default Branch: master
• Last Pushed: 2024-01-07T00:34:04.000Z (over 2 years ago)
• Last Synced: 2025-02-10T00:17:00.525Z (over 1 year ago)
• Language: Jupyter Notebook
• Size: 25 MB
• Stars: 2
• Watchers: 3
• Forks: 1
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

          
Tensorflow Cats vs Dogs classifier

---

# Ipython notebook new Mobilenet model

Using TF Keras to transfer learn + fine-tune a MobileNetV2 model on our own dataset of dogs/cats.

We get around 93% accuracy on the test dataset.

[Ipython notebook training testing model](dogcat.ipynb)

![predictions](predictions.png)

# Old model

We build our own dataset from existing flicker images of cats and dogs, and then train a tensorflow neural network to classify cats and dogs.

Run using `./cats_vs_dogs.py`

Currently a single layer NN, no successful learning yet

```

usage: cats_vs_dogs.py [-h] [--cat_dir CAT_DIR] [--dog_dir DOG_DIR]

                       [--num_steps NUM_STEPS]

optional arguments:

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

  --cat_dir CAT_DIR     Directory for storing input cat images

  --dog_dir DOG_DIR     Directory for storing input dog images

  --num_steps NUM_STEPS

                        Number of steps to train model

```

# Building a dataset

`resize_images.py` contains a script to resize all passed in images into

64x64 grayscale pngs named `####.png` monotonically increasing in the 

specified output folder

```

Usage: resize_images.py [options] image1 [image2 ...]

Options:

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

  -o OUTPUT_FOLDER, --output_folder=OUTPUT_FOLDER

                        Output folder to save resized images to

  -n MAX_N, --max_number=MAX_N

                        Maximum number of images to process

  -d, --dryrun          Do a dry run (no processing/saving)

```