Image shadow removal model and construction method, device and application thereof

The invention provides an image shadow removal model construction method and device and application, and the method comprises the following steps: obtaining a training sample, and carrying out the preprocessing of the training sample, and obtaining a pre-screened shadow image and a shadow mask; encoding by using a first encoder and a second encoder to obtain a first encoding result and a second encoding result; adding position information to the first coding result and the second coding result by using a cross-region Transform layer, and then sending the first coding result and the second coding result into a region perception cross attention layer to obtain a shadow feature map; and using a RefineNet network to perform coding and decoding by taking the original image, the pre-screened shadow image and the shadow feature map as input to obtain a shadow removal result map corresponding to the original image. According to the scheme, the cross-region Transform layer and the region perception cross attention lay