I ran a yolov7 model to detect trash and I want to know how to interpret the confusion matrix. Image consists of two classes: either the object that is detected is trash or not trash. But there seems to be a new class called Background FN along the row and Background FP along the column.
I understand that FN and FP mean false positive and false negative. But I assumed that for a 2 class problem there's gonna be two rows and two columns with the typical TP, TN, FP, FN values. Now there's these extra probability values and I don't understand where they come from. The dataset I used comes from here: https://universe.roboflow.com/nam-nhat/trash-dvdrr?utm_source=augmented-startups&utm_medium=video&utm_campaign=yolov7-tutorial This confusion matrix is created automatically with other curves like precision vs confidence curve and precision vs recall curves once the yolov7 model is built. Now I want to predict the accuracy of the model as well as the false alarm rate, so I thought knowing the TP, TN, FP, FN values will be helpful but I don't understand this background class. I'm unable to post image of confusion matrix so I'm making a table of it.
<table>
<tr>
<th></th>
<th>Not Trash</th>
<th>Trash</th>
<th>Background FP</th>
</tr>
<tr>
<td>Not Trash</td>
<td>0.77</td>
<td>0.02</td>
<td>0.27</td>
</tr>
<tr>
<td>Trash</td>
<td>0.02</td>
<td>0.97</td>
<td>0.73</td>
</tr>
<tr>
<td>Background FN</td>
<td>0.21</td>
<td>0.02</td>
<td></td>
</tr>
</table>Object detection is a bit different; understand it as follows.
Background FP refers to background objects that do not belong to either of the classes but detected as one of them.
Background FN refers to Trash or Non-trash objects missed by the detector and considered as some other background objects.