DEV Community

Cover image for FruitsClassifier on Fruits 360
Quoc Bao
Quoc Bao

Posted on

FruitsClassifier on Fruits 360

On this project, I build a classifier on the Fruits 360 dataset using Pytorch. I use a pretrained model and transfer learning, as well as do hyper-parameter search to help increase the accuracy.

 

✍️ Documentation:

- Full Training on Kaggle | GoogleNet | 94% Acc

- Efficient Training with Pytorch Lightning

- GitHub Page

 

📰 Dataset structure

train/test
    |___apple_braeburn_1
        |___r0_0.jpg
        |___r0_10.jpg
            ...
    |___carrot_1
        |___r0_0.jpg
        |___r0_10.jpg
        ...
Enter fullscreen mode Exit fullscreen mode

 

Large batch visualization

batch

 

Separate visualization

se

 

Single fruit visualization

singbatch

 

Single fruit visualization with transforms

singtr

 

Separate visualization with transforms

setr

 

Batch visualization with transforms

batchtr

 

Pixel Intensity

pixel

 

Visualize the number of classes

class

 

Visualize loss and accuracy

lossacc

 

Visualize model predictions: green(correct) and red(incorrect)

pred

Top comments (0)