Edge Impulse Experts / Lightweight-Face-Detection Public
Target: Qualcomm Dragonwing RB3 Gen 2 Development Kit Clone this project

Test data

Set thresholds
Model variants
No additional model variants available Test 'Unoptimized (float32)' model
Set the 'expected outcome' for each sample to the desired outcome to automatically score the impulse.

Apply filters

Sample name Expected outcome Precision score Result
bounding_boxes face, face 72%
Show classification Show in data acquisition Show in feature explorer
Filter yielded no results, click to adjust the filter

Model testing output

Results

Model version:
mAP@50
1.00

Metrics for Pretrained learn

Metric Value
mAP 0.15
mAP@[IoU=50] 1.00
mAP@[IoU=75] 0.00
mAP@[area=small] -1.00
mAP@[area=medium] -1.00
mAP@[area=large] 0.15
Recall@[max_detections=1] 0.10
Recall@[max_detections=10] 0.15
Recall@[max_detections=100] 0.15
Recall@[area=small] -1.00
Recall@[area=medium] -1.00
Recall@[area=large] 0.15
Precision score (legacy) 72.5%

Confusion matrix

Loading...

Feature explorer

Loading...
Help & documentation Get help on the forum Send feedback Contact sales
Join us
Terms of Service Privacy policy