润色In the coding stage high-level semantic features are extracted from the rail image by convolution operation In the decoding phase progressive upsampling is used to restore the size of the image and

During the coding stage, the rail image is processed through convolution operation to extract high-level semantic features. These features capture the underlying information and patterns present in the image.

In the decoding phase, progressive upsampling techniques are employed to increase the size of the image while preserving the extracted features. This upsampling process helps to restore the image to its original dimensions.

Furthermore, based on the extracted features, a predictive segmentation mask is generated. This mask predicts the different segments or regions within the image, helping in the identification and separation of various objects or areas of interest