Skip to content

Latest commit

 

History

History
 
 

dog_emotions

Folders and files

NameName
Last commit message
Last commit date

parent directory

..
README.md
README.md
 
 
clean_images.ipynb
clean_images.ipynb
 
 
dog_labels.txt
dog_labels.txt
 
 
download_images.ipynb
download_images.ipynb
 
 
imagenet_classes.txt
imagenet_classes.txt
 
 
prediction.ipynb
prediction.ipynb
 
 
remove_images.ipynb
remove_images.ipynb
 
 
settings.py
settings.py
 
 

README.md

Project Overview

In this project, we'll figure out what emotion a dog is feeling based on a picture. We'll use deep learning to analyze the image and predict the emotion. We'll start out by downloading the images from Flickr. Then we'll clean the images using PIL. We'll remove any images that don't have a dog in them using timm and a pretrained deep learning model. Then we'll then train a neural network to make predictions about what emotion a dog is feeling. We'll end by evaluating how to improve the model.

We'll use Pytorch and torchvision to train the model.

At the end, you'll have a trained deep learning model that can predict dog emotions.

Project Steps

  • Download images from Flickr
  • Clean images using PIL
  • Remove any images that don't have a dog in them
  • Train a deep learning model to predict dog emotions
  • Evaluate the model

Code

You can find the code for this project here

File overview:

  • download_images.ipynb - code to download the images from Flickr.
  • clean_images.ipynb - clean the images using PIL.
  • remove_images.ipynb - remove any images that don't include a dog.
  • prediction.ipynb - train a neural network to predict if a dog is in an image.
  • settings.py - has constants that are needed in other files.

Prerequisites

To complete this project, you'll need to have a good understanding of:

  • Python syntax, including functions, if statements, and data structures
  • Data cleaning
  • Pandas syntax
  • Using Jupyter notebook

You'll also need to know the basics of the command line and machine learning.

Please make sure you've completed these Dataquest courses (or know the material) before trying this project:

Local Setup

Installation

To follow this project, please install the following locally:

  • JupyerLab
  • Python 3.8+
  • Python packages
    • pandas
    • numpy
    • timm
    • torch
    • torchvision
    • Pillow
    • flickrapi

Data

We'll download the image data using Flickr.

  • If you don't want to get an API key and download the data, you can find the images here.
  • If you don't want to clean the images, and want to jump into deep learning, you can download cleaned images here.
  • The model with pretrained weights is here