Select which labels to recognize and set a threshold for each:
The confidence threshold determines how certain the model needs to be to detect an object. Here's how to use these settings:
- To detect more objects: Lower the threshold (e.g., 0.3). This will show more detections but might include some false positives.
- For more accurate detections: Increase the threshold (e.g., 0.7). This will show fewer detections but with higher confidence.
- To hide a category: Simply uncheck its checkbox. The model will ignore this category completely.
- Default threshold: 0.5 (50% confidence) is a good starting point for most cases.