What tools to use to analyze yolov3 with squeezenet network miss detection?
4 Ansichten (letzte 30 Tage)
Ältere Kommentare anzeigen
gil
am 23 Dez. 2021
Beantwortet: Srivardhan Gadila
am 27 Dez. 2021
We trained yolov3ObjectDetector as in the example
https://www.mathworks.com/help/vision/ug/object-detection-using-yolo-v3-deep-learning.html and the loss was close to 0.
While running [bboxes,scores,labels] = detect(detector.detector, resized, "Threshold", 0) with the above yolov3ObjectDetector, we get no results even for the training images.
We tried analyzing the network by using activations https://www.mathworks.com/help/deeplearning/ref/seriesnetwork.activations.html but it expects DAGNetwork object and the yolov3ObjectDetector is of type dlnetwork.
Can we convert dlnetwork to DAGNetwork and use the activations?
Any other tool that can help?
0 Kommentare
Akzeptierte Antwort
Srivardhan Gadila
am 27 Dez. 2021
The YOLO v3 network in the yolov3ObjectDetector is stored as a dlnetwork (Deep Learning Toolbox) object. Whereas the function activations expects input network to be a SeriesNetwork or a DAGNetwork object. Hence the function activations does not work for dlnetwork object.
But in case of dlnetwork, you can get the output of any required layer by using the following syntax of predict function on dlnetwork object:
You can refer to the documentation of predict (https://www.mathworks.com/help/deeplearning/ref/dlnetwork.predict.html) for more information.
0 Kommentare
Weitere Antworten (0)
Siehe auch
Community Treasure Hunt
Find the treasures in MATLAB Central and discover how the community can help you!
Start Hunting!