metadata
datasets:
- garythung/trashnet
language:
- en
metrics:
- accuracy
library_name: keras
This repository contains a sample work to classify garbage type based on resized images on this repository.
There are 2 models available:
- trash-classification-no-aug.keras
- trash-classification-aug.keras
The trash-classification-no-aug.keras model trained with no (or minimum) data augmentation:
datagen = ImageDataGenerator(
rescale=1./255,
validation_split=0.2
)
While the trash-classification-aug.keras model trained with more data augmentation works in the dataset:
# With data augmentation
datagen = ImageDataGenerator(
rescale=1./255,
validation_split=0.2,
width_shift_range=0.1,
height_shift_range=0.1,
horizontal_flip=True
)
The models trained with Tensorflow Functional API by using this approach:
Conv --> BatchNorm --> Conv --> BatchNorm --> MaxPooling (3x)
For the detailed description about the training process and models' performace, you can visit this Github repository.