|
|
General System Parameters |
|
|
| verboseLogging |
The level of logging of the work of neural networks. |
- 0 – logging off; |
2 |
 |
|
|
- 1 – logging errors; |
|
|
|
|
- 2 – logging warnings; |
|
|
|
|
- 3 – logging information; |
|
|
|
|
- 4 – logging debug information; |
|
|
|
|
- 5 – logging all. |
|
|
| betaMode |
Beta testing mode for any new functionality added as experiment. |
- 0 – off; |
0 |
 |
|
|
- 1 – on. |
|
|
| defaultDetectorType |
Default detector type. |
- FaceDetV1 |
FaceDetV1 |
|
|
|
- FaceDetV2 |
|
|
|
|
- FaceDetV3 |
|
|
|
|
Detector HumanDetector::Settings |
|
|
| ScoreThreshold |
Human detection accuracy threshold. Increasing the threshold reduces the number of false positive detections but increases the number of false negative detection (misses). |
0.000…1.000 |
0.31 |
 |
| RedetectScoreThreshold |
Threshold for estimation of human redetection accuracy. Increasing the threshold reduces the number of false negative detection results. |
0.000…1.000 |
0.12 |
|
| NMSThreshold |
Threshold value for non-maximum suppression. The higher the threshold, the higher the probability of detecting a human blocked by objects. Increasing the threshold will increase the number of false positive detections. |
0.000…1.000 |
0.4 |
|
| RedetectNMSThreshold |
The value of the non-maximum suppression threshold for human redetection. The higher the threshold, the higher the probability of repeated detection of a human blocked by objects. Increasing the threshold will increase the number of false positive detections. |
0.000…1.000 |
0.4 |
|
| imageSize |
The size of the image in pixels along the larger side, on which the human is detected. Increasing the size will allow better detection of humans the background, but it will increase the processing time. |
320...1280 |
640 |
|
| nms |
A parameter that affects the accuracy of localization of the human on the frame. The higher the accuracy, the lower the speed. |
- best – simplified mode of operation; |
mean |
|
|
|
- mean – enhanced precision mode. |
|
|
| RedetectNMS |
Parameter affecting the accuracy of human localization on the frame. Higher accuracy, reduces performance. |
- best – simplified mode of operation; |
mean |
|
|
|
- mean – enhanced precision mode. |
|
|
| landmarks17Threshold |
Threshold for the accuracy of determining 17 body landmarks. The parameter is used to filter and exclude false marks. |
0.000…1.000 |
0.2 |
|
| RedetectTensorSize |
Determining the size of the tensor (the primary data structure for the neural network) for the neural network of the redetect. This parameter is provided by the Research team, has one default value and must not be changed by users. |
- |
110 |
 |
| RedetectHumanTargetSize |
The size of a human crop for a re-detector. This parameter is provided by the Research team, has one default value and must not be changed by users. |
- |
85 |
|