Update README.md
Browse files
README.md
CHANGED
|
@@ -151,9 +151,9 @@ We use the following _statistical-based metrics_:
|
|
| 151 |
- **Image Recall** measures the percentage of correct images in the multimodal answer relative to the total number of images in the ground truth, evaluating whether the answer effectively includes useful image information.
|
| 152 |
It is computed as:
|
| 153 |
|
| 154 |
-
\[
|
| 155 |
\text{Image Recall} = \frac{\text{True Positives}}{\text{True Positives} + \text{False Negatives}}
|
| 156 |
-
\]
|
| 157 |
|
| 158 |
where False Negatives are the images in the ground truth that were omitted in the generated multimodal answer.
|
| 159 |
|
|
|
|
| 151 |
- **Image Recall** measures the percentage of correct images in the multimodal answer relative to the total number of images in the ground truth, evaluating whether the answer effectively includes useful image information.
|
| 152 |
It is computed as:
|
| 153 |
|
| 154 |
+
![\[
|
| 155 |
\text{Image Recall} = \frac{\text{True Positives}}{\text{True Positives} + \text{False Negatives}}
|
| 156 |
+
\]](https://cdn-uploads.huggingface.co/production/uploads/67571051d39ac252085797ca/dJCpcB_oeZDmjn0AAGbex.png)
|
| 157 |
|
| 158 |
where False Negatives are the images in the ground truth that were omitted in the generated multimodal answer.
|
| 159 |
|