This code detects faces in live video using YOLO algorithm and Harr-Cascade classifier.
It uses Keras framework along with OpenCV and Pillow library.
- TensorFlow
- Keras
- OpenCV
- Pillow
- YOLOv3 Weights: Can be downloaded from here or here.
- Haar-Cascade Classifier: Can be found here on the OpenCV GitHub repo, and can be downloaded from here.
-
Harr-Cascade classifier by itself is not that accurate and often returns false-positives (As seen in the following examples):
-
Detection of full body in image/frame by YOLOv3 (Very Accurate):
-
Implementing Haar-Cascade classifier inside regions detected by YOLOv3 gives much more accurate results:
