Alvtron

public static class EnumerableSplitExtensions
{
    /// <summary>
    /// Splits the enumerable into two separate sequences based on predicate.
    /// </summary>
    /// <remarks>
    /// This is more efficient than using two Where()-filters (from LINQ) because the source is only enumerated one time. Testing reveals an approximate 30% speed increase.
    /// </remarks>
    /// <typeparam name="TSource">The type of the elements in the enumerable.</typeparam>
    /// <param name="source">The source of elements.</param>

Alvtron

public static class EnumerableCountExtensions
{
    /// <summary>
    /// Check if the number of elements in the source is equal to the specified count.
    /// </summary>
    /// <remarks>
    /// This will never exhaust the source past the necessary number of elements.
    /// This is in almost all circumstances faster than .Count() == x, where x is an arbitrary <see cref="int"/>.<br/>
    /// If the source is a sequence that implements <see cref="ICollection{TSource}"/>, the Count-property is used instead.
    /// </remarks>

Alvtron

Stratified dataset split in PyTorch

When working with imbalanced data for machine learning tasks in PyTorch, and simple random split might not be able to partly divide classes that are not well represented. Resulting sample splits might not portray the real-world population, leading to poor predictive peformance in the resulting model.

Therefore, I have created a simple function for conducting a stratified split with random shuffling, similar to that of StratifiedShuffleSplit from scikit-learn (https://scikit-learn.org/stable/modules/generated/sklearn.model_selection.StratifiedShuffleSplit.html)

import random
import math
import torch.utils.data

Alvtron

Reflect

Reflect values between a minimum and maximum. I have written both an iterative and recursive version of the algorithm.

Iterative (fastest)

import math

def reflect(value, min_value, max_value):
    span = max_value - min_value
    if value < min_value:

Alvtron

Download 411 Bob Ross Paintings

Download all 411 Bob Ross paintings from the website Two Inch Brush (https://www.twoinchbrush.com) with Python.

This is an update to a similar solution made by Jeffrey Thompson (https://gist.github.com/jeffThompson/6d4c45f89dc907925775972e72d9cf00). I had problems running his code with Python 3, so I wrote my own code that essentially does the same thing.

Code

import requests

NUM_IMAGES = 411