DRUNet: A Method for Infrared Point Target Detection

Deep learning is widely used in vision tasks, but feature extraction of IR small targets difficult due to the inconspicuous contours and lack color information. This paper proposes a new convolutional neural network–based (CNN-based) method for target detection called DRUNet. The algorithm divided into two parts: network prediction head. For targets, which are accurately label, Gaussian soft labels added supervise training process make converge faster. We use simplified object keypoint similarity evaluate accuracy by ratio distance radius inner tangent circle box fair evaluating model inference speed after GPU preheating. experimental results show that our proposed performs better when compared with commonly algorithms field detection. size 10.5 M, test reaches 133 FPS under RTX3090 platform.