Skip to content

Palmgrenoskari/Birds_CNN_PyTorch

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

20 Commits
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Birds_CNN_PyTorch

Learning project for school.
Idea is to present your work in detail to other students in the course.

CNN Architecture

EfficientNet-B4 (pretrained)

Data

High quality birds dataset from kaggle: https://www.kaggle.com/datasets/gpiosenka/100-bird-species/data

  • 525 bird species -> reduced to 184 (Only bird names starting with A,B or C)
  • 84635 training img -> reduced to 29274
    • Each species has at least 130 training imgs
    • ~80% Male and 20% female, this is a shortcoming in the dataset
  • 2625 test & validation imgs (~3% each, very low IMO) -> reduced to 920
  • 224x224x3 JPG

Approach

Data is prepared, we can go almost straight into modeling and evaluation.
Plan is to make a nice well explained Jupyter Notebook that can be presented live (except the training of course)

  • Save the model
  • Explain everything
  • Visualizations

Note: Again, this is a learning project. Never coded PyTorch myself before this.
Materials I've used so far:

About

Learning project for school.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published