Update README.md

README.md

@@ -10,67 +10,103 @@ tags:
   - central asian food dataset
   - object-detection
 paper: https://doi.org/10.1038/s41598-025-95770-9 
-homepage: https://github.com/IS2AI/Central_Asian_Food_Scenes_Dataset  # optional
+homepage: https://github.com/IS2AI/Central_Asian_Food_Scenes_Dataset  
 ---
 
 
 
-# Central_Asian_Food_Scenes_Dataset
-In this work, we propose the first Central Asia Food Scenes Dataset that contains 21,306 images with 69,856 instances across 239 food classes. To make sure that the dataset contains various food items, we took as a benchmark the ontology of Global Individual Food Tool developed by Food and Agriculture Organization (FAO) together with the World Health Organization (WHO) [1]. The dataset contains food items across 18 coarse classes: vegetables, baked flower-based products, cooked dishes, fruits, herbs, meat dishes, desserts, salads, sauces, drinks, dairy, fast-food, soups, sides, nuts, pickled and fermented food, egg product, and cereals. 
+# Central Asian Food Scenes Dataset
 
-The dataset contains open source web-scraped images from the search engines (15,939 images) (i.e., Google, YouTube, and Yandex) and our own collected food images from everyday life (2,324 images). To additionally extend the number of instances of the underrepresented classes, we have scraped open-source videos and extracted frames at a rate one frame per second (3,043 images). The dataset has been checked and cleaned for duplicates using the Python Hash Image library. Furthermore, we have also filtered out images less than 30 kB in size and replaced them by performing additional iterative data scraping and duplicate check to make sure the high quality of the dataset.
+In this work, we propose the first Central Asia Food Scenes Dataset that contains **21,306 images** with **69,856 instances** across **239 food classes**. To make sure that the dataset contains various food items, we took as a benchmark the ontology of Global Individual Food Tool developed by Food and Agriculture Organization (FAO) together with the World Health Organization (WHO) [1]. The dataset contains food items across **18 coarse classes**: 
+🍅 Vegetables • 🥖 Baked goods • 🍲 Cooked dishes • 🍎 Fruits • 🌿 Herbs • 🍖 Meat dishes • 🍰 Desserts • 🥗 Salads • 🥫 Sauces • 🥤 Drinks • 🧀 Dairy • 🍔 Fast food • 🍜 Soups • 🍟 Sides • 🥜 Nuts • 🥒 Pickled & fermented • 🥚 Egg products • 🌾 Cereals  
+
+The images come from:  
+- 🤖 **15,939** web-scraped images (i.e., Google, YouTube, Yandex)  
+- 📸 **2,324** everyday-life photos  
+- 🎥 **3,043** video frames (1 fps)  
+
+The dataset has been checked and cleaned for duplicates using the Python Hash Image library. Furthermore, we have also filtered out images less than 30 kB in size and replaced them by performing additional iterative data scraping and duplicate check to make sure the **high quality** of the dataset.
+
+---
+
+## 📊 Class Balance
 
 The dataset is unbalaced. The statistics across high-level 18 classes is shown on Figure below.
 
-<img src="https://raw.githubusercontent.com/IS2AI/Central_Asian_Food_Scenes_Dataset/main/figures/statistics.png" width="800" height="900">
+<div align="center">
+  <img
+    src="https://raw.githubusercontent.com/IS2AI/Central_Asian_Food_Scenes_Dataset/main/figures/statistics.png"
+    alt="Class distribution"
+    width="700"
+  >
+</div>
+
+---
 
-Sample Food scenes annotated images are shown below. Figures illustrate annotated food scenes samples based on our annotation rules that we have followed to create the dataset: the liquid objects such as beverages and soups are annotated together with the dish itself (see upper image), solid food items are annotated without the plate, in case one class is located on top of another class the annotations are made as shown on lower image; in case one class is obscured by another class and the rest of the background class is not visible we highlight only the visible part (see `Salad leaves' class annotation on the lower image. 
+## 🖼️ Annotation Examples
 
-<img src="https://raw.githubusercontent.com/IS2AI/Central_Asian_Food_Scenes_Dataset/main/figures/protocol_1.png" width="800" height="650">
-<img src="https://raw.githubusercontent.com/IS2AI/Central_Asian_Food_Scenes_Dataset/main/figures/protocol_2.png" width="800" height="680">
+Sample Food scenes annotated images are shown below. Figures illustrate annotated food scenes samples based on our annotation rules that we have followed to create the dataset: the liquid objects such as beverages and soups are annotated together with the dish itself (see upper image), solid food items are annotated without the plate, in case one class is located on top of another class the annotations are made as shown on lower image; in case one class is obscured by another class and the rest of the background class is not visible we highlight only the visible part (see `Salad leaves' class annotation on the lower image). 
 
+<div align="center">
+  <img
+    src="https://raw.githubusercontent.com/IS2AI/Central_Asian_Food_Scenes_Dataset/main/figures/protocol_1.png"
+    alt="Protocol 1"
+    width="700"
+  /><br><br>
+  <img
+    src="https://raw.githubusercontent.com/IS2AI/Central_Asian_Food_Scenes_Dataset/main/figures/protocol_2.png"
+    alt="Protocol 2"
+    width="700"
+  >
+</div>
 
+---
 
-# Dataset split across the folds
+## 🔀 Dataset Splits
 
 |Fold| Train| Valid| Test| 
 |----|-----------------|-----------------|--------------------|
 |images|17,046|2,084|2,176|
 |instances|55,422|7,062|7,381|
 
+---
+
+## 📈 Training Results
 
-# Training results
+Training results of different versions of the YOLOv8 model on the Central Asian Food Scenes Dataset, model sizes, and inference times.
 
-To illustrate the performance of different classification models on CAFD we have trained different models. We used the largest publicly available fine-grained dataset Food1K [[1]](#1) that contains 1,000 food classes to evaluate the performance of classifier with the 1,042 food categories.
+<div align="center">
+  <img
+    src="https://raw.githubusercontent.com/IS2AI/Central_Asian_Food_Scenes_Dataset/main/figures/table_results.png"
+    alt="Training results"
+    width="800"
+  >
+</div>
 
-<img src="https://raw.githubusercontent.com/IS2AI/Central_Asian_Food_Scenes_Dataset/main/figures/table_results.png" width="1200" height="290">
+---
 
-# Pre-trained models
+## 🛠️ Pre-trained Models
 
 Pre-trained weights of the YOLOv8 model can be downloaded using these links:
 
-## YOLOv8n: 
-https://issai.nu.edu.kz/wp-content/themes/issai-new/data/models/CAFSD/yolov8n.pt
-
-## YOLOv8s: 
-https://issai.nu.edu.kz/wp-content/themes/issai-new/data/models/CAFSD/yolov8s.pt
+- **YOLOv8n**: https://issai.nu.edu.kz/wp-content/themes/issai-new/data/models/CAFSD/yolov8n.pt
+- **YOLOv8s**: https://issai.nu.edu.kz/wp-content/themes/issai-new/data/models/CAFSD/yolov8s.pt
+- **YOLOv8m**: https://issai.nu.edu.kz/wp-content/themes/issai-new/data/models/CAFSD/yolov8m.pt  
+- **YOLOv8l**: https://issai.nu.edu.kz/wp-content/themes/issai-new/data/models/CAFSD/yolov8l.pt  
+- **YOLOv8x**: https://issai.nu.edu.kz/wp-content/themes/issai-new/data/models/CAFSD/yolov8x.pt  
 
-## YOLOv8m: 
-https://issai.nu.edu.kz/wp-content/themes/issai-new/data/models/CAFSD/yolov8m.pt
+---
 
-## YOLOv8l: 
-https://issai.nu.edu.kz/wp-content/themes/issai-new/data/models/CAFSD/yolov8l.pt
+## 📖 References
 
-## YOLOv8x: 
-https://issai.nu.edu.kz/wp-content/themes/issai-new/data/models/CAFSD/yolov8x.pt
+[1] European Commission: Impact assessment on measures addressing food waste to complete swd (2014) 207 regarding the review of EU waste management target (2014). Accessed 5-22-2024.
 
+---
 
-## References
-<a id="1">[1]</a> 
-European Commission: Impact assessment on measures addressing food waste to complete swd (2014) 207 regarding the review of EU waste management target (2014). Accessed 5-22-2024.
+## 📝 Citation
 
+In case of using our dataset and/or pre-trained models, please cite our work:
 
-# In case of using our dataset and/or pre-trained models, please cite our work:
 ```bibtex
 @article{Karabay2025,
   author    = {Karabay, Aknur and Varol, Huseyin Atakan and Chan, Mei Yen},
@@ -85,5 +121,7 @@ European Commission: Impact assessment on measures addressing food waste to comp
   issn      = {2045-2322}
 }
 ```
-[📖 View the publication](https://doi.org/10.1038/s41598-025-95770-9)
+
+[📚 View the publication](https://doi.org/10.1038/s41598-025-95770-9)
+