ISPR midterms taken during the course 2020/2021
Implement the convolution of a set of edge detection filters with an image and apply it to one face image and one tree image of your choice from the dataset. Implement Roberts, Prewitt and Sobel filters (see here, Section 5.2, for a reference) and compare the results (it is sufficient to do it visually). You should not use the library functions for performing the convolution or to generate the Sobel filter. Implement your own and show the code!
Implement from scratch an RBM and apply it to DSET3. The RBM should be implemented fully by you (both CD-1 training and inference steps) but you are free to use library functions for the rest (e.g. image loading and management, etc.).
-
Train an RBM with 100 hidden neurons (single layer) on the MNIST data (use the training set split provided by the website).
-
Use the trained RBM to encode all the images using the corresponding activation of the hidden neurons.
-
Train a simple classifier (e.g. any simple classifier in scikit) to recognize the MNIST digits using as inputs their encoding obtained at step 2. Use the standard training/test split. Show the resulting confusion matrices (training and test) in your presentation.
(Alternative to step 3, optional) Step 3 can as well be realized by placing a softmax layer after the RBM: if you are up for it, feel free to solve the assignment this way.
Implement your own convolutional network, deciding how many layers, the type of layers and how they are interleaved, the type of pooling, the use of residual connections, etc. Discuss why you made each choice a provide performance results of your CNN on CIFAR-10.
Now that your network is trained, you might try an adversaria attack to it. Try the simple Fast Gradient Sign method, generating one (or more) adversarial examples starting from one (or more) CIFAR-10 test images. It is up to you to decide if you want to implement the attack on your own or use one of the available libraries (e.g. foolbox, CleverHans, ...). Display the original image, the adversarial noise and the final adversarial example.
Paper: Defences against adversarial attacks – arxiv.org/pdf/1702.04267.pdf