Abstract: Highlights•A wavelet pooling-based multiscale contextual fusion network, WMCF-Net, is proposed.•A novel MCF block is proposed to capture fine and coarse details from key regions.•A comparative analysis with state-of-the-art CNNs is performed.•The results on four datasets demonstrate the superiority of WMCF-Net.•The ablation experiments and visualization results further indicate its effectiveness.
Loading