Update README.md
Browse files
README.md
CHANGED
@@ -27,7 +27,7 @@ The dataset contains 20,578 RGB images of 120 dog breeds. Images are labeled as
|
|
27 |
- sitting (3038 images)
|
28 |
- lying down (7090 images)
|
29 |
- undefined (6307 images)
|
30 |
-
|
31 |
Images have varying resolutions, with 50% between 361x333 and 500x453 pixels.
|
32 |
|
33 |
#### Data Collection and Processing
|
@@ -38,7 +38,7 @@ This dataset is an adaption of from the [Stanford Dog Dataset](http://vision.sta
|
|
38 |
|
39 |
The dataset has a class imbalance, with nearly 2x as many "lying down" images compared to "sitting". Indistinguishable poses were labeled as "undefined", with most being close-up portraits. This may limit the ability to handle such images.
|
40 |
|
41 |
-
**Recommendations**: When using this dataset, be aware of the class imbalance and consider oversampling or augmentation techniques
|
42 |
|
43 |
## Citation
|
44 |
|
|
|
27 |
- sitting (3038 images)
|
28 |
- lying down (7090 images)
|
29 |
- undefined (6307 images)
|
30 |
+
|
31 |
Images have varying resolutions, with 50% between 361x333 and 500x453 pixels.
|
32 |
|
33 |
#### Data Collection and Processing
|
|
|
38 |
|
39 |
The dataset has a class imbalance, with nearly 2x as many "lying down" images compared to "sitting". Indistinguishable poses were labeled as "undefined", with most being close-up portraits. This may limit the ability to handle such images.
|
40 |
|
41 |
+
**Recommendations**: When using this dataset, be aware of the class imbalance and consider oversampling or augmentation techniques..
|
42 |
|
43 |
## Citation
|
44 |
|