This code is 3d-CNN model that can predict environmental value. Firstly, I built a model that can create a lot of building randomly in the area which we gave it. Secondly, I recorded the output value, such as visibility, average hour sun and radiation, by inputting different seed to random function. Thirdly, I created a matrix that bound the whole model, and the process was likely to recording model in voxel way, so that I could get the model information by the matrix. Forthly, after preparing the 1000 samples for training set, I used it to train 3d-CNN. Finally, this 3d-CNN model can be trained to predict the environmental value. Just saving the 3d-CNN model and executing it by cpython, and it can work in grasshopper. The 3d-CNN is like the eyes of computer that can scan and read the status of the model, so I don’t need to define the value which represnts the 3d model.