/cs231n

Stanford CS231n Spring 2019 assignments(with myself solution), and tutorials(English and Chinese). Completed both tensorflow and pytorch version

Primary LanguageJupyter Notebook

CS231N Convolutional Neural Networks for Visual Recognition

Assignments

This repository provides CS231N assignment and answers (not official)

Assignment #1: Image Classification, kNN, SVM, Softmax, Neural Network

Assignment #2: Fully-Connected Nets, Batch Normalization, Dropout, Convolutional Nets

Assignment #3: Image Captioning with Vanilla RNNs, Image Captioning with LSTMs, Network Visualization, Style Transfer, Generative Adversarial Networks

Docs

Enviroments

  • OS: Windows 10

  • Tensorflow: 2.0-beta1

  • Pytorch: 1.2.0

Tips

For the dataset needed in assignment, run .sh files to get dataset if you are Linux user, otherwise for Windows user, here are two tips:

  1. For examples, open get_datasets.sh file, you shall see these codes:
# Get CIFAR10
wget http://www.cs.toronto.edu/~kriz/cifar-10-python.tar.gz
tar -xzvf cifar-10-python.tar.gz
rm cifar-10-python.tar.gz

Manually copy the url http://www.cs.toronto.edu/~kriz/cifar-10-python.tar.gz to browser, and the browser will automatically start downloading.

  1. For those urls can not open in the browser, copy the link, and open Xunlei(迅雷) to create a new task, fill in this url, and Xunlei will start downloading.

Problems

You may meet some problems as below :

1. Import Error: no module named 'past'

pip install future

2. ImportError: cannot import name 'imread'

imread is deprecated! imread is deprecated in SciPy 1.0.0, and will be removed in 1.2.0. Use imageio.imread instead.

  • Solvement1: degrade scipy to 1.2.0 or less
  • Solvement2: use from imageio import imread instead.

3. ImportError: cannot import name 'imresize'

imresize is deprecated! imresize is deprecated in SciPy 1.0.0, and will be removed in 1.2.0. Use skimage.transform.resize instead.

  • Solvement1: degrade scipy to 1.2.0 or less
  • Solvement2: use from skimage.transform import resize as imresize instead.

4. Object arrays cannot be loaded when allow_pickle=False

open file cs231n/data_utils.py, revise f = np.load(imagenet_fn) to f = np.load(imagenet_fn,allow_pickle=True) in function load_imagenet_val(num=None) , line 256.

5. TypeError: 'numpy.float64' object is not iterable

which may occur in assignment3/StyleTransfer.ipynb, open cs231n/image_utils.py, revise img = imresize(img, scale_factor) to img = imresize(img, new_shape) in function load_image(filename, size=None), line 90.

6. AssertionError:

you may encounter it in assignment3/Generative_Adversarial_Networks_PyTorch.ipynb, block [7], replace assert np.any(np_z < 0.0) and np.any(np_z > 0.0) with assert not np.any(np_z < 0.0) and np.any(np_z > 0.0)