Zindi Challenge Preprcosseing

Author

Eric Wanjau

1 Description

Note

This is jus a preprocessing script for:

Spot the Mask Challenge by#ZindiWeekendz

Link to challenge:

https://zindi.africa/competitions/zindiweekendz-learning-spot-the-mask-challenge/data

The objective of this challenge is to create an image classification machine learning model to accurately predict the likelihood that an image contains a person wearing a face mask, or not. The total dataset contains 1,800+ images of people either wearing masks or not.

You will need to use train_labels.csv to split the data into the train and test sets.

1 indicates an image with a person with a mask, 0 indicates a person without a mask.

2 Splitting data into train and test sets

# Load required packages
library(tidyverse)
library(here)

# Import train and test labels
train_labels <- read_csv("train_labels.csv")
test_files <- read_csv("SampleSubmission.csv") %>% 
  pull(image)

# Obtain train masked and nonmasked image files
# This can be done in a function: group_by + nest + map? if more categories?

masked <- train_labels %>% 
  filter(target == 1) %>% 
  pull(image)

nonmasked <- train_labels %>% 
  filter(target == 0) %>% 
  pull(image)

# How many masked images
length(masked)

# How many nonmasked images
length(nonmasked)

# How many test files
length(test_files)

3 Place the images into their own folders

# Image file names
all_files <- list.files("images")

# Create train images folder
dir.create("train_images/masked")
dir.create("train_images/nonmasked")

# dir.exists("images")

# Create test folder
dir.create("test_images")

# Copy masked files from `images` folder into `train/masked` folder
invisible(file.copy(from = here("images", masked),
          to = here("train_images", "masked"),
          overwrite = TRUE))

# Copy unmasked files from `images` folder to `train/nonmasked` folder
invisible(file.copy(from = here("images", nonmasked),
          to = here("train_images", "nonmasked"),
          overwrite = TRUE))

# Copy test files into test folder
invisible(file.copy(from = here("images", test_files),
          to = here("test_images"),
          overwrite = TRUE))

# Alternative solution for test files
# What are the test files
# Discard train files from all_files
test_files <- discard(all_files, all_files %in% train_labels$image)

# Copy test files into test folder
invisible(file.copy(from = here("images", test_files),
          to = here("test_images"),
          overwrite = TRUE))

--- title: "Zindi Challenge Preprcosseing" author: "Eric Wanjau" toc: true format: html: number-sections: true toc-location: left code-tools: true code-fold: false code-link: true editor: visual execute: eval: false warning: false message: false --- ## Description ::: callout-note This is jus a preprocessing script for: **Spot the Mask Challenge by[ #ZindiWeekendz](http://www.zindi.africa/)** Link to challenge: <https://zindi.africa/competitions/zindiweekendz-learning-spot-the-mask-challenge/data> ::: The objective of this challenge is to create an image classification machine learning model to accurately predict the likelihood that an image contains a person wearing a face mask, or not. The total dataset contains 1,800+ images of people either wearing masks or not. You will need to use train_labels.csv to split the data into the train and test sets. > `1` indicates an image with a person with a `mask`, `0` indicates a person `without a mask`. ## Splitting data into train and test sets ```{r} # Load required packages library(tidyverse) library(here) # Import train and test labels train_labels <- read_csv("train_labels.csv") test_files <- read_csv("SampleSubmission.csv") %>% pull(image) # Obtain train masked and nonmasked image files # This can be done in a function: group_by + nest + map? if more categories? masked <- train_labels %>% filter(target == 1) %>% pull(image) nonmasked <- train_labels %>% filter(target == 0) %>% pull(image) # How many masked images length(masked) # How many nonmasked images length(nonmasked) # How many test files length(test_files) ``` ## Place the images into their own folders ```{r} # Image file names all_files <- list.files("images") ``` ```{r} # Create train images folder dir.create("train_images/masked") dir.create("train_images/nonmasked") # dir.exists("images") # Create test folder dir.create("test_images") # Copy masked files from `images` folder into `train/masked` folder invisible(file.copy(from = here("images", masked), to = here("train_images", "masked"), overwrite = TRUE)) # Copy unmasked files from `images` folder to `train/nonmasked` folder invisible(file.copy(from = here("images", nonmasked), to = here("train_images", "nonmasked"), overwrite = TRUE)) # Copy test files into test folder invisible(file.copy(from = here("images", test_files), to = here("test_images"), overwrite = TRUE)) ``` ```{r} # Alternative solution for test files # What are the test files # Discard train files from all_files test_files <- discard(all_files, all_files %in% train_labels$image) # Copy test files into test folder invisible(file.copy(from = here("images", test_files), to = here("test_images"), overwrite = TRUE)) ```