The goal of this project is to classify images using convolutional neural networks. We have investigated three different approaches that are presented in the following notebooks:

• scratch.ipynb: a CNN is trained from scratch. The CIFAR database is used for this purpose.
• pre-trained.ipynb: a photograph of an object is fed to a deep learning model and the top 3 most likely classes for the image and their probabilities are retuned.
• fine-tuning.ipynb: a pre-trained deep learning model is fine-tuned on a new set of classes.

The ResNet50 model is used in this project. This model is trained on a subset of the ImageNet database, which is used in the ImageNet Large-Scale Visual Recognition Competition (ILSVRC).

We use the Keras deep learning library with TensorFlow as a backend for this project. It is worth noting that state-of-the-art models (VGG, Inception, ResNet, ...) with their implementation and pre-trained weights on common dataset like ImageNet or CIFAR are easily accessible with Keras.

Open a notebook and follow the instructions. For instance:

jupyter notebook pre-trained.ipynb