#Intel Landscape Image Classification | A Network Architecture comparison
##Introduction:
The intel landscape image classification dataset is quite similar to the previously explored ICFAR10 dataset. It consists of roughly 25k 150x150px images of 6 different categories.
The dataset was initially published (https://datahack.analyticsvidhya.com/) by Intel to host a Image classification Challenge.
The one used here can be found on kaggle to offer easy integration into colab:
https://www.kaggle.com/puneet6060/intel-image-classification
###Goals:
The main goal of this project was to first implement different architectures learnt throughout the course and beyond, to get a better feel for the different limitations of the architectures:
- Logistic Regression
- Feed Forward NN
- Simple Convolutional Neural Network
- Residual Neural Network (Resnet 9)
- Residual Neural Network Variation (Resnet18)
After benchmarking the different architectures the best one is to be picked and augmented with different technics to achive a accuracy as close as possible to the one achieved in the competition (96%)