Assignments of TensorFlow 2 for Deep Learning Specialization program by Imperial London College
COURSE 1: Getting Started with TensorFlow2
- 01-convolutional-neural-network: It is first coding assignment in the project. It includes additional iterations such as increasing number of epochs, adding Dropout layer (dropout = 0.5), building smaller model, and changing padding type. MNIST dataset is used.
- 02-model-validation: It is second coding assignment in the course. Regularization techniques and callbacks methods of Keras library are studied. Iris dataset is used for the assignment.
- 03-Capstone-project: It is final project of the course. SVHN dataset is used to train an MLP model and a CNN model using their guidelines. Previously studied techniques are implemented to the project.
COURSE 2: Customizing Models with Tensorflow2
- 001-transfer-learning-tutorial: The notebook includes introduction to Keras Functional API and adjusting layers manually.
- 002-transfer-learning-assignment: Its first assignment of the course. There is a comparison between CNN and Transfer Learning. Dogs vs. Cats dataset is used in the project.
- 003-data-pipeline-tutorial: Data pipeline tutorial for image processing using Keras.
- 004-data-pipeline-assignment: Second week assignment for image preprocessing & using tf.Keras datasets.
- 005-sequence-modelling-tutorial.ipynb: Sequential modelling & text processing are practiced.
- 006-sequence-modelling-assignment: Third week assignment is completed using Shakespeare.
- 007-model-subclassing-tutorial: There are model subclassing tutorial & related notes. These practices include custom optimization and training processes.
- 008-model-subclassing-assignment: It is forth week assignment that covers Residual Network implementation using Fashion MNIST dataset.
- 009-capstone-project: It is final assignment of the course. A custom model which has AutoEncoder structure is developed for English to German translation using English to German dataset. Its html format is available here.