https://github.com/ssiarhei115/celebrities-classification-w-pytorch

https://github.com/ssiarhei115/celebrities-classification-w-pytorch

Last synced: 14 days ago
JSON representation

• Host: GitHub
• URL: https://github.com/ssiarhei115/celebrities-classification-w-pytorch
• Owner: ssiarhei115
• Created: 2024-07-23T15:18:17.000Z (6 months ago)
• Default Branch: master
• Last Pushed: 2024-07-23T18:30:41.000Z (6 months ago)
• Last Synced: 2024-07-24T18:04:35.112Z (6 months ago)
• Language: Jupyter Notebook
• Size: 7.16 MB
• Stars: 0
• Watchers: 1
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

        
# Multiclass classification using a pretrained neural network

## Main goal

Make multiclass celebrities classification model using pretrained Neural Network

## Data description

Primary dataset consists of 2 folders - train (3000 images), valid (914 images); each of them contains 5 folders with pictures of celebrities: Elon Mask, Bill Gates, Jeff Bezos, Mark Zuckerberg, Steve Jobs. Folder 'test' contains one image for prediction test purpose, it was downloaded from the Internet.

## Metric

Main metric used - accuracy.

## Summary

Pretrained ResNet34 model with following modifications was used to achieve the goal:

* Last fully connected layer was replaced with new one with 5 outputs according to the quantity of classes expected.

* Only last layer's weights were updated during the training procedure



As a result accuracy value more than 90% was achived on validation subset.

## Libraries & tools used

* see the requirements