This project was developed for the Statistical Methods for Machine Learning course (academic year 2021-2022). The project addresses the task of binary image classification by considering 25000 images of cats and dogs. In this scenario, the proposed approach consists in training different Convolutional Neural Networks, which are more suitable in case of image classification.
- Python 3.10.11
- Tensorflow version: 2.9.2
- OpenCV
- Project CatsDogs CNNs.jpynb: file containing data preprocessing and CNNs execution.
- Report containing more theoretical details about the implemented CNNs and the relative training results.
This link below connects to the shared CatsDogs folder containing three subfolders (Cats, Dogs and Pickles). As written in the GoogleColab file, if you want to skip all the data preprocessing phases, I suggest you to directly use or download the files contained in Pickles (images_grey.pickle, labels_grey.pickle) and run the code from the section named 'Normalizing pixel values' present in the GoogleColab file. By doing so, you will be ready for running the CNNs models! More details are in GoogleColab file. Since Colab is connected to Google Drive, I suggest to use this folder to speed up the process. https://drive.google.com/drive/folders/1rCcfqym_UIUSDBr2_uOUBwt-f7FUr7y7?usp=sharing