Machine Learning - Convolutional Neural Network
Waste Classification
This Project Focusses on the multi classification of household trash items. It aims to create a CNN to predict what type of trash something is and classify the image into the respective category.
Data is from Kaggle
Used Data Augmentation to create more images
The Model:
Conv2D Layer 1: Conv2D(32, (3, 3))
MaxPooling2D Layer 1: MaxPooling2D(pool_size=(2, 2))
Conv2D Layer 2: Conv2D(64, (3, 3))
MaxPooling2D Layer 2: MaxPooling2D(pool_size=(2, 2))
Conv2D Layer 3: Conv2D(128, (3, 3))
MaxPooling2D Layer 3: MaxPooling2D(pool_size=(2, 2))
Flatten Layer: Flatten()
Dense Layer 1: Dense(128)
Output Layer: Dense(9, activation='softmax')
Results:
Training Accuaracy - 73%
Cross Validation Accuracy - 52%
Found new data set for test set:
Test Accuarcy - 26%