Deep Learning Model for Satellite Image Classification

Apr 27, 2016
Code Follow
Photo by rawpixel on Unsplash
  1. In this project, we developed and trained two deep-learning models on a satellite image dataset from Kaggle.
  2. The first model is a regular CNN-based model with 17 layers and uses a cross-entropy loss function to learn the weights/kernels. (RegNet17 model)
  3. The second model is based on the residual architecture added to conventional CNNs. This model has 17 layers and uses a cross-entropy loss function for learning the weights/kernels. (ResNet17 model)
  4. Finally, both models were compared for speed and accuracy.
  5. Although RegNet17 is more accurate, ResNet17 has low inference time, which is vital for real-time object classification.
Deep Learning
Abhay Kumara Sri Krishna Nandiraju
Abhay Kumara Sri Krishna Nandiraju
MSIS:ML at University of Arizona