How to train a classifier for 2D geometric data?

I have 2D geometric data which when rendered results into floorplans like the one shown in this image

Now for each segment in the figure above I have the end points data ie the 2 points that when connected makes up the line. Now, I want to train a classifier to classify different lines into their respective classes. I have labeled the data already.

My confusion is with regards to how should I go about making the model to also take into consideration the relationships between the lines as multiple lines with same dimension could belong to different classes, or should I not worry about this as a model would pickup this relationship automatically.

I did not understand your data properly. Do you want to train without rendering into an image? I think it’s better if you have rendered images as train data and their corresponding labels with different colors, so that it becomes a image segmentation problem. Correct me if my understanding of the problem is wrong.