Stanford CS231n: Convolutional Neural Networks for Visual Recognition

Solutions for CS231n Assignments

assignment1:

• kNN Classifier
• SVM Classifier
• Softmax Classifier
• Two Layer Neural Net (network implementation in neural_net.py)
• Feature Representations: experiment with higher level representations (Histogram of Gradients)

assignment2:

• Fully Connected Nets: Modular layer design in layers.py and fc_net.py
• BatchNorm
• Dropout
• CNN: CNN, Max Pooling, Spatial Batchnorm, GroupNorm implementations

assignment3:

• RNN_Captioning: implements a Vanilla RNN for image captioning MS-COCO dataset. Layer implementations are in rnn_layers
• LSTM_Captioning: implements a LSTM for the same task
• Network Visualization: uses a pretrained SqueezeNet model to compute gradients with respect to images and product saliency maps and fooling images in Tensorflow
• Style Transfer: implements style transfer in Tensorflow
• Generative Adversarial Networks: implements vanilla GAN, Least Squares GAN, and DCGAN in Tensorflow