Learning model architecture for image data semantic segmentation

A system comprising:

• one or more processors;
• and
• memory communicatively coupled to the one or more processors, the memory storing computer-executable modules executable by the one or more processors that, when executed by the one or more processors, perform associated operations, the computer-executable modules comprising:
• a convolutional layer module configured to:
• perform down-sampling convolution upon image features outputting to a higher convolutional layer module;
• perform convolution upon same resolution image features;
• and
• receive lower resolution image features from a higher convolutional layer;
• and
• a feature fusing module configured to fuse the lower resolution image features received from the higher convolutional layer with the same resolution image features convoluted at a same convolutional layer.

A computer-readable storage medium storing computer-readable instructions executable by one or more processors, that when executed by the one or more processors, cause the one or more processors to perform operations comprising:

• performing, at a convolutional layer of a learning model, down-sampling convolution upon image features outputting to a higher convolutional layer;
• performing, at the convolutional layer, convolution upon same resolution image features;
• receiving, at the convolutional layer, lower resolution image features from the higher convolutional layer;
• and
• fusing, at the convolutional layer, the lower resolution image features received from the higher convolutional layer with the same resolution features convoluted at the same convolutional layer.