Update README.md
Browse files
README.md
CHANGED
|
@@ -1,8 +1,20 @@
|
|
| 1 |
-
|
| 2 |
-
|
| 3 |
-
|
| 4 |
-
|
| 5 |
-
|
| 6 |
-
|
| 7 |
-
|
| 8 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
### 📊 Dataset Statistics
|
| 2 |
+
|
| 3 |
+
• Total Images: 1,227 (predominantly .jpg format)
|
| 4 |
+
• Total Annotations: 1,441 (.txt files in YOLO format)
|
| 5 |
+
• Total Bounding Boxes: 2,219 (across the matched images)
|
| 6 |
+
• Missing Images: There are 214 annotation files (.txt) that do not have a corresponding image file. They should be filtered out before training models.
|
| 7 |
+
|
| 8 |
+
### 🏷️ Class Mapping & Distribution (Matched Pairs)
|
| 9 |
+
|
| 10 |
+
The YOLO annotations map to the following severity classes:
|
| 11 |
+
|
| 12 |
+
• Class 0 (First-Degree Burn): 872 bounding boxes (39.3%)
|
| 13 |
+
• Class 1 (Second-Degree Burn): 992 bounding boxes (44.7%)
|
| 14 |
+
• Class 2 (Third-Degree Burn): 355 bounding boxes (16.0%)
|
| 15 |
+
|
| 16 |
+
### 🛠️ Key Sections in the README.md
|
| 17 |
+
|
| 18 |
+
• YOLO Format Reference: Quick explanation of the space-separated coordinates structure.
|
| 19 |
+
• Usage Guidance: A Python snippet to automatically clean up/filter out the 214 unmatched .txt files.
|
| 20 |
+
• Citation Information: A standard BibTeX entry for the associated paper ("Detection of Different Degrees of Skin Burn using YOLOv3", Baid et al., 2020) for easy reference.
|