1 // Ceres Solver - A fast non-linear least squares minimizer 2 // Copyright 2010, 2011, 2012 Google Inc. All rights reserved. 3 // http://code.google.com/p/ceres-solver/ 4 // 5 // Redistribution and use in source and binary forms, with or without 6 // modification, are permitted provided that the following conditions are met: 7 // 8 // * Redistributions of source code must retain the above copyright notice, 9 // this list of conditions and the following disclaimer. 10 // * Redistributions in binary form must reproduce the above copyright notice, 11 // this list of conditions and the following disclaimer in the documentation 12 // and/or other materials provided with the distribution. 13 // * Neither the name of Google Inc. nor the names of its contributors may be 14 // used to endorse or promote products derived from this software without 15 // specific prior written permission. 16 // 17 // THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" 18 // AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 19 // IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 20 // ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE 21 // LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 22 // CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 23 // SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 24 // INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 25 // CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 26 // ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 27 // POSSIBILITY OF SUCH DAMAGE. 28 // 29 // Author: keir (a] google.com (Keir Mierle) 30 31 syntax = "proto2"; 32 33 package ceres.internal; 34 35 message BlockProto { 36 // The span of the block. 37 optional int32 size = 1; 38 39 // Position along the row or column (depending on storage orientation). 40 optional int32 position = 2; 41 } 42 43 message CellProto { 44 // Column or row block id as appropriate. 45 optional int32 block_id = 1; 46 47 // Position in the values array the cell is located. Each cell is stored as a 48 // row-major chunk inside the values array. 49 optional int32 position = 2; 50 } 51 52 // A single row or column, depending on the matrix type. 53 message CompressedRowProto { 54 optional BlockProto block = 2; 55 repeated CellProto cells = 1; 56 } 57 58 message BlockStructureProto { 59 repeated BlockProto cols = 1; 60 repeated CompressedRowProto rows = 2; 61 } 62 63 // A block sparse matrix, either in column major or row major format. 64 message BlockSparseMatrixProto { 65 optional int64 num_rows = 2; 66 optional int64 num_cols = 3; 67 optional int64 num_nonzeros = 4; 68 repeated double values = 1 [packed=true]; 69 70 optional BlockStructureProto block_structure = 5; 71 } 72 73 message TripletSparseMatrixProto { 74 optional int64 num_rows = 4; 75 optional int64 num_cols = 5; 76 optional int64 num_nonzeros = 6; 77 78 // The data is stored as three arrays. For each i, values(i) is stored at the 79 // location (rows(i), cols(i)). If the there are multiple entries with the 80 // same (rows(i), cols(i)), the values entries corresponding to them are 81 // summed up. 82 repeated int64 rows = 1 [packed=true]; 83 repeated int64 cols = 2 [packed=true]; 84 repeated double values = 3 [packed=true]; 85 } 86 87 message CompressedRowSparseMatrixProto { 88 optional int64 num_rows = 4; 89 optional int64 num_cols = 5; 90 91 repeated int64 rows = 1 [packed=true]; 92 repeated int64 cols = 2 [packed=true]; 93 repeated double values = 3 [packed=true]; 94 } 95 96 message DenseSparseMatrixProto { 97 optional int64 num_rows = 1; 98 optional int64 num_cols = 2; 99 100 // Entries are stored in row-major order. 101 repeated double values = 3 [packed=true]; 102 } 103 104 // A sparse matrix. It is a union; only one field is permitted. If new sparse 105 // implementations are added, update this proto accordingly. 106 message SparseMatrixProto { 107 optional TripletSparseMatrixProto triplet_matrix = 1; 108 optional BlockSparseMatrixProto block_matrix = 2; 109 optional CompressedRowSparseMatrixProto compressed_row_matrix = 3; 110 optional DenseSparseMatrixProto dense_matrix = 4; 111 } 112 113 // A linear least squares problem. 114 // 115 // Given a matrix A, an optional diagonal matrix D as a vector, and a vector b, 116 // the proto represents the following linear least squares problem. 117 // 118 // | A | x = | b | 119 // | D | | 0 | 120 // 121 // If D is empty, then the problem is considered to be 122 // 123 // A x = b 124 // 125 // The desired solution for the problem is the vector x that solves the 126 // following optimization problem: 127 // 128 // arg min_x ||Ax - b||^2 + ||Dx||^2 129 // 130 // If x is present, then it is the expected solution to the 131 // problem. The dimensions of A, b, x, and D should be consistent. 132 message LinearLeastSquaresProblemProto { 133 optional SparseMatrixProto a = 1; 134 repeated double b = 2 [packed=true]; 135 repeated double d = 3 [packed=true]; 136 repeated double x = 4 [packed=true]; 137 // If the problem is of SfM type, i.e it has a generalized 138 // bi-partite structure, then num_eliminate_blocks is the number of 139 // column blocks that are to eliminated in the formation of the 140 // Schur complement. For more details see 141 // explicit_schur_complement_solver.h. 142 optional int32 num_eliminate_blocks = 5; 143 } 144
