Method for training shallow convolutional neural networks for infrared target detection using a two-phase learning strategy
Disclosed is a method for training shallow convolutional neural networks for infrared target detection using a two-phase learning strategy that can converge to satisfactory detection performance, even with scale-invariance capability. In the first step, the aim is to ensure that only filters in the convolutional layer produce semantic features that serve the problem of target detection. L2-norm (Euclidian norm) is used as loss function for the stable training of semantic filters obtained from the convolutional layers. In the next step, only the decision layers are trained by transferring the weight values in the convolutional layers completely and freezing the learning rate. In this step, unlike the first, the L1-norm (mean-absolute-deviation) loss function is used.
1. A method for training shallow convolutional neural networks for infrared target detection using a two-phase learning strategy, comprising:
training filters in a convolutional layer with an L2-norm loss function to produce semantic features and trained filters, and
training fully connected decision layers with an L1-norm loss function by using the trained filters,
wherein the fully connected decision layers are kept smaller than layers without two-phase learning to prevent overfitting while training the filters in a first time.