In　this　paper,　the　Back　Propagation　(BP)　network　and　Radial　Basis　Function　(RBF)　neural　network　are　employed　to　recognize　and　reconstruct　3D　freeform　surface　from　2D　freehand　sketch.　Some　tests　and　comparison　experiments　have　been　made　to　evaluate　the　performance　for　the　reconstruction　of　freeform　surfaces　of　both　networks　using　simulation　data.　The　experimental　results　show　that　both　BP　and　RBF　based　freeform　surface　reconstruction　methods　are　feasible;　and　the　RBF　network　performed　better.　The　RBF　average　point　error　between　the　reconstructed　3D　surface　data　and　the　desired　3D　surface　data　is　less　than　0.05　over　all　our　75　test　sample　data.