Our Dataset

We used Caltech101 dataset to create a new dataset for our project. We divided 2159 sample images into 2 classes (planes V.S. non planes) and used 1015 images to retrain the CNN, 674 images for validation and 470 images for testing our newly trained model. (validation VS testing)

L. Fei-Fei, R. Fergus and P. Perona. Learning generative visual models from few training examples: an incremental Bayesian approach tested on 101 object categories. IEEE. CVPR 2004, Workshop on Generative-Model Based Vision. 2004