Exemplary embodiments include a method for edge correction in pattern recognition, the method including receiving a design layout, receiving a sample plan based on the design layout, receiving user-generated edge input and generating a recipe output from the design layout, the sample plan and the user-generated edge input. The incorporation of the edge input results in SEM recipes that are much more successful in recognizing patterns that have tendency to deviate in appearance from design by, for example, moderate to severe sidewall angle.