Fig. 19From: Intelligent radar HRRP target recognition based on CNN-BERT modelAttention map of each head of deep multi-head self-attention layer; a–e attention map of 5 different heads, f average attention mapBack to article page