Training parameter adaptive optimization method for unmanned aerial vehicle target identification online learning
The invention discloses a training parameter adaptive optimization method for unmanned aerial vehicle target recognition online learning, and the method comprises the steps: employing an image which is actually shot by an unmanned aerial vehicle as a subsequent online learning sample, and carrying o...
Gespeichert in:
Hauptverfasser: | , , , , , , , |
---|---|
Format: | Patent |
Sprache: | chi ; eng |
Schlagworte: | |
Online-Zugang: | Volltext bestellen |
Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
Zusammenfassung: | The invention discloses a training parameter adaptive optimization method for unmanned aerial vehicle target recognition online learning, and the method comprises the steps: employing an image which is actually shot by an unmanned aerial vehicle as a subsequent online learning sample, and carrying out the continuous training of an unmanned aerial vehicle target recognition model; in the continuous training process, the online learning samples are divided according to the set batch number, and the divided batch samples are adopted to perform iterative training on the unmanned aerial vehicle target recognition model; calculating the convergence speed of the unmanned aerial vehicle target recognition model according to the calculation time and the loss value obtained after each iteration; and comparing the calculated convergence rate of the current iteration period with a historical optimal convergence rate, and updating the optimal convergence rate or batch number according to a comparison result. According to |
---|