Indonesian Food Image
Description
The dataset is a diverse collection of colored images featuring various Indonesian foods. These images display a range of backgrounds where the foods are presented, including restaurants, roadside stalls, and traditional eateries. The collected images also include variations in the background objects, such as tableware like plates, bowls, forks, spoons, and knives, or additional condiments like soy sauce, chili sauce, and lime. This stage utilizes a Google Chrome extension called Fatkun to perform scraping of image files on the browser page. The browser page used is Google Images, and the search keyword is the names of the Indonesian food types. The Fatkun extension can save all the images on that browser page into a chosen directory. The Indonesian foods are divided into ten classes of images, namely; meatball soup (bakso), seasoned duck (bebek betutu), mixed vegetable salad with peanut sauce (gado-gado), jackfruit stew (gudeg), fried rice (nasi goreng), fishcake (pempek), black soup (rawon), spicy meat dish (rendang), skewered and grilled meat (sate), and traditional soup (soto).
Files
Steps to reproduce
Here are the steps to reproduce the dataset gathered. 1. Search on Google Images using the keyword of the food: This step involves using a web browser to open Google Images and entering the keyword of the food you want to gather images for. For example, if you want to gather images of "pizza," you would search for "pizza" on Google Images. 2. Gather all the images in the page using the Google Chrome extension Fatkun: Fatkun is a Google Chrome extension that allows you to bulk download images from a webpage. After performing the Google Images search, you would use the Fatkun extension to download all the images displayed on the page. This extension saves you time and effort by automating the image download process. 3. Save it to a specified directory: The downloaded images using the Fatkun extension will be saved to a specified directory on your computer. It's essential to organize the images neatly into separate folders for each class (food category) to facilitate the next steps of the process. 4. Repeat until you have 10 classes: To create a diverse dataset, you would repeat steps 1 to 3 for each of the ten food classes you want to include in your dataset. This way, you ensure that you have a variety of food images from different classes. 5. Check them manually if the images gathered are food images with the specified class: After gathering images for each class, it's crucial to manually review the images to ensure they belong to the correct class and are relevant to your food image dataset. This step is crucial for quality control and to avoid including irrelevant images in your dataset. 6. Split the folders into training and testing using the Split Folders library from the Python programming language: The final step involves splitting the collected images into two subsets: training and testing sets. This split is essential for training and evaluating machine learning models effectively. The Split Folders library in Python can help you achieve this by randomly dividing the images into separate folders for training and testing. Where in this dataset 70% is allocated for training and 30% is allocated for testing.