1. Bash the kaggleAPI.sh script. You will need to have a Kaggle account, installation and API credentials (learn more here). You can either use a terminal from Linux/Mac OS or RStudio (learn more here).
2. Install all packages listed on top of the script. EBImage is deposited in BioConductor, all other packages in CRAN.
3. Run script.R. The TensorFlow instalation defaults to CPU; alternatively you can use NVIDIA GPUs (if you have them) which run substantially faster.

This work is partly based on:

• Image processing - Image Classification in R: MXNet from Shikun Li
• CNN architecture - Keras CIFAR-10 CNN example

Enjoy, all feedback is welcome!

Francisco