https://github.com/kelvinjuliusarmandoh/dog_vision

Transfer learning for predicting dog breed images classification
https://github.com/kelvinjuliusarmandoh/dog_vision

computer-vision deep-learning machine-learning multi-class-classification transfer-learning

Last synced: 11 days ago
JSON representation

Transfer learning for predicting dog breed images classification

• Host: GitHub
• URL: https://github.com/kelvinjuliusarmandoh/dog_vision
• Owner: kelvinjuliusarmandoh
• Created: 2023-06-18T07:55:19.000Z (over 1 year ago)
• Default Branch: main
• Last Pushed: 2023-06-18T09:46:24.000Z (over 1 year ago)
• Last Synced: 2024-11-05T20:18:55.585Z (about 2 months ago)
• Topics: computer-vision, deep-learning, machine-learning, multi-class-classification, transfer-learning
• Homepage:
• Size: 12.6 MB
• Stars: 0
• Watchers: 2
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

        
# Dog Images Classification using Transfer Learning

This notebook builds a end-to-end multiclass image classifier using TensorFlow 2.0 and TensorFlow Hub.

![golden](https://github.com/KelvinJulius/dog_vision/assets/108222785/25954cba-fdd6-4f1a-9742-4094d549e9fe)

## Problem

Given a  dog picture. Guess what kind of dog it is?

## About the Data

Get the data for dog vision from Kaggle. Here is the link:

https://www.kaggle.com/c/dog-breed-identification/overview

## Evaluation

The evaluation metrics is probabilities output (multiclass output) from softmax activation.

https://www.kaggle.com/competitions/dog-breed-identification/overview/evaluation

## Features

* We're using deep learning/transfer learning because it's probably best way for images (unstructured data).

* There are 120 unique kind of dogs in this data.

* There are 10,000+ images in the training set (Images with labels)

* There are around 10,000+ images in the test set (Images have no labels)

  

## Result

Succesfully to handle overfitting using data augmentation. Get 84% and 82% for loss training score and validating score.

## Workflow of this project

* Preparing the data (Loading from Kaggle, preprocessing the images, and split the data into 3 sets (training, validating, testing))

* Choosing a model and Fitting the model (Using Tensorflow Hub and prevent the overfitting using `earlystopping` callback)

* Monitoring the performance of loss training and loss validating (using tensorboard callback)

* Improve the model (Using Image Augmentation for reducing the overfitting model)

* Save the model