Food Image Classification Dataset
This dataset provides images and information on 10 types of foods commonly consumed by Koreans.
- Labeling Type: Food
- Data Format: Image
- Data Type: Synthetic Data
Main Product
Data Quantity (Samples)
Total Price
$ 7,900
(VAT Included)Looking for custom-made dataset or researcher-accessible data? Please contact us for inquiries.
About Dataset
1) Data Introduction
• This dataset provides images and information on 10 types of foods commonly consumed by Koreans. The data is classified into 10 food classes: ""Ramen"", ""Gimbap"", ""Curry"", ""Army Stew (Budae Jjigae)"", ""Corn Soup (Corn Soup)"", ""Jajangmyeon (Black Bean Noodles)"", ""Boiled Dumplings (Mul Mandu)"", ""Fried Dumplings (Gun Mandu)"", ""White Rice (Baekmi)"", and ""Red Bean Porridge (Patjuk)".2) Data Utilization
(1) Food Data has characteristics that: • The dataset includes images of each food item, which are useful for understanding and analyzing the characteristics of these foods. • It provides information that helps in understanding the diverse types and features of Korean cuisine. (2) Food Data can be used to: • Food analysis and trend research: Professionals in the food industry can utilize this data to analyze current food trends and develop new menus. • Development of recommendation systems: It can be used to develop personalized food recommendation systems on e-commerce platforms and food delivery apps, providing tailored recommendations based on customer preferences. • Machine learning and computer vision research: It can be utilized to train and evaluate machine learning models in areas such as food image classification and recognition. • Cultural research: Analyzing various Korean foods can be utilized for cultural research and educational materials. • Digital content creation: It can be used to create digital content for blogs, social media, online magazines, etc., to introduce and promote various Korean foods."Meta Data
Good
Outstanding
Data Samples 4
Utility
| Downstream Classification (▲) | KID (▼) | One Class Classification (▼) | |
|---|---|---|---|
| Total | 0 | 0 | 0 |
| Suitability | OK | OK | OK |
The higher the value, the better (▲)
Model Performance
Downstream classification accuracy is an indicator used to evaluate the usefulness of synthetic data. It measures whether synthetic data performs similarly to real data. The method involves training the same model separately on real data and synthetic data, and then comparing the accuracies of the two models. Interpretation: A high accuracy rate means that the model trained on synthetic data performs similarly to the one trained on real data, indicating that the synthetic data is of high quality and well represents the real data.
The closer to zero or the lower the value, the better (▼)
Quality
KID (Kernel Inception Distance) is a metric used to evaluate the similarity between generated images and real images. It compares the differences between the two sample distributions using Kernel Mean Embedding, without assuming a normal distribution. Interpretation: A lower KID score suggests that generated images are more similar to real images, with a score close to 0 being ideal. Specifically, a score below 0.01 indicates very high similarity.
Privacy
| LPIPS (▲) | SSIM (▼) | |
|---|---|---|
| Total | 0 | 0 |
| Suitability | OK | OK |
The higher the value, the better (▲)
Perceptual Similarity
Learned Perceptual Image Patch Similarity (LPIPS) is a metric used to measure the visual similarity between two images by utilizing neural networks to extract key features and calculate the distance between them. High LPIPS value: Indicates high similarity between images, raising the risk of information leakage. Low LPIPS value: Suggests that synthetic images are perceptually different from real images, indicating a lower risk of sensitive information leakage.
The closer to zero or the lower the value, the better (▼)
Structural Similarity
The Structural Similarity Index Measure (SSIM) is a metric used to assess the similarity between two images. It is primarily used to compare the quality of a restored or compressed image with the original image. SSIM measures visual similarity by considering brightness, contrast, and structure.
High SSIM value (0.9 or above): Indicates that the synthetic image is very similar to the real image, which may increase the risk of information leakage.
Low SSIM value (0.6 or below): Indicates low similarity and reduced risk of leakage.
Premium Report Information
Premium dataset sample
