- Generated on Fri Apr 20 2012 16:46:40 for libeblearn by
1.7.4
|
libeblearn
|
00001 /*************************************************************************** 00002 * Copyright (C) 2010 by Pierre Sermanet * 00003 * pierre.sermanet@gmail.com * 00004 * All rights reserved. 00005 * 00006 * Redistribution and use in source and binary forms, with or without 00007 * modification, are permitted provided that the following conditions are met: 00008 * * Redistributions of source code must retain the above copyright 00009 * notice, this list of conditions and the following disclaimer. 00010 * * Redistributions in binary form must reproduce the above copyright 00011 * notice, this list of conditions and the following disclaimer in the 00012 * documentation and/or other materials provided with the distribution. 00013 * * Redistribution under a license not approved by the Open Source 00014 * Initiative (http://www.opensource.org) must display the 00015 * following acknowledgement in all advertising material: 00016 * This product includes software developed at the Courant 00017 * Institute of Mathematical Sciences (http://cims.nyu.edu). 00018 * * The names of the authors may not be used to endorse or promote products 00019 * derived from this software without specific prior written permission. 00020 * 00021 * THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED 00022 * WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED 00023 * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE 00024 * DISCLAIMED. IN NO EVENT SHALL ThE AUTHORS BE LIABLE FOR ANY 00025 * DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES 00026 * (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; 00027 * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND 00028 * ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT 00029 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS 00030 * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. 00031 ***************************************************************************/ 00032 00033 #ifndef DETECTOR_H_ 00034 #define DETECTOR_H_ 00035 00036 #include "libidx.h" 00037 #include "ebl_states.h" 00038 #include "ebl_arch.h" 00039 #include "ebl_answer.h" 00040 #include "ebl_merge.h" 00041 #include "ebl_preprocessing.h" 00042 #include "bbox.h" 00043 #include "nms.h" 00044 00045 using namespace std; 00046 00047 namespace ebl { 00048 00050 // detector 00051 00061 enum t_scaling { MANUAL = 0, SCALES = 1, NSCALES = 2, SCALES_STEP = 3, 00062 ORIGINAL = 4, NETWORK = 5, SCALES_STEP_UP = 6 }; 00063 00064 template <typename T, class Tstate = fstate_idx<T> > 00065 class detector { 00066 public: 00067 00069 // constructors 00070 00084 detector(module_1_1<T,Tstate> &thenet, vector<string> &labels, 00085 answer_module<T,T,T,Tstate> *answer = NULL, 00086 resizepp_module<T,Tstate> *resize = NULL, 00087 const char *background = NULL, 00088 std::ostream &out = std::cout, std::ostream &err = std::cerr, 00089 bool adapt_scales = false); 00090 00092 virtual ~detector(); 00093 00095 // configuration 00096 00098 void set_scaling_original(); 00100 void set_scaling_type(t_scaling type); 00102 void set_resolutions(const midxdim &scales); 00104 void set_resolutions(const vector<double> &factors); 00106 void set_resolution(double factor); 00110 void set_resolutions(int resolutions); 00114 void set_resolutions(idx<uint> &resolutions); 00117 void set_zpads(float hzpad, float wzpad); 00125 void set_resolutions(double scales_steps, double max_scale = 1.0, 00126 double min_scale = 1.0); 00128 int get_class_id(const string &name); 00130 void set_bgclass(const char *bg = NULL); 00135 bool set_mask_class(const char *mask); 00137 void set_silent(); 00142 void set_min_resolution(uint min_size); 00147 void set_max_resolution(uint max_size); 00149 void set_raw_thresholds(vector<float> &t); 00150 00153 void set_nms(t_nms type = nms_overlap, float pre_threshold = 0.0, 00154 float post_threshold = 0.0, 00155 float pre_hfact = 1.0, float pre_wfact = 1.0, 00156 float post_hfact = 1.0, float post_wfact = 1.0, 00157 float woverh = 1.0, float max_overlap = 1.0, 00158 float max_hcenter_dist = 0.0, float max_wcenter_dist = 0.0, 00159 float vote_max_overlap = 1.0,float vote_max_hcenter_dist = 0.0, 00160 float vote_max_wcenter_dist = 0.0); 00161 00163 void set_scaler_mode(bool set); 00165 void set_smoothing(uint type); 00166 00173 void set_mem_optimization(Tstate &in, Tstate &out, 00174 bool keep_inputs = false); 00176 void set_netdim(idxdim &d); 00182 void set_outputs_dumping(const char *name); 00184 void set_bboxes_off(); 00186 vector<string>& get_labels(); 00188 void set_ignore_outsiders(); 00192 void set_corners_inference(uint type); 00196 void set_bbox_decision(uint type); 00197 void set_bbox_scalings(mfidxdim &scalings); 00198 00200 // execution 00201 00207 template <class Tin> bboxes& fprop(idx<Tin> &img, const char *fname = NULL); 00209 void fprop_nms(bboxes &in, bboxes &out); 00212 vector<idx<T> >& get_originals(); 00215 midx<T> get_preprocessed(const bbox &b); 00223 svector<midx<T> >& get_preprocessed(bboxes &out, uint n = 0, 00224 bool diverse = false, 00225 uint pre_diverse_max = 100); 00234 svector<midx<T> >& get_preprocessed(bboxes &in, bboxes &out, uint n = 0, 00235 bool diverse = false, 00236 uint pre_diverse_max = 100); 00239 idx<T> get_mask(string &classname); 00241 uint get_total_saved(); 00248 string& set_save(const string &directory, uint nmax = 0, 00249 bool diversity = false); 00251 void init(idxdim &dinput, const char *frame_name = NULL); 00252 00253 protected: 00254 // scales methods ////////////////////////////////////////////////////////// 00255 00262 void compute_scales(midxdim &scales, idxdim &netdim, idxdim &mindim, 00263 idxdim &maxdim, idxdim &indim, t_scaling type, 00264 uint nscales, double scales_step, 00265 const char *frame_name = NULL); 00268 void compute_resolutions(midxdim &scales, 00269 idxdim &mindim, idxdim &maxdim, uint nscales); 00272 void compute_resolutions(midxdim &scales, 00273 idxdim &indim, vector<double> &scale_factors); 00278 void compute_resolutions(midxdim &scales, idxdim &mindim, 00279 idxdim &maxdim, double scales_step); 00284 void compute_resolutions_up(midxdim &scales, idxdim &indim, 00285 idxdim &mindim, idxdim &maxdim, 00286 double scales_step); 00290 void validate_resolutions(); 00291 00292 // bboxes operations /////////////////////////////////////////////////////// 00293 00295 void smooth_outputs(); 00298 void update_merge_alignment(); 00301 void get_corners(mstate<Tstate> &outputs); 00304 void extract_bboxes(T threshold, bboxes &bboxes); 00307 void save_bboxes(bboxes &bboxes, const string &dir, 00308 const char *frame_name = NULL); 00312 void add_class(const char *name); 00313 00314 // processing methods ////////////////////////////////////////////////////// 00315 00320 template <class Tin> void prepare(idx<Tin> &img, const char *fname = NULL); 00326 void prepare_scale(uint i); 00328 void multi_res_fprop(); 00329 00330 // member variables //////////////////////////////////////////////////////// 00331 protected: 00332 module_1_1<T,Tstate> &thenet; 00333 resizepp_module<T,Tstate> *resizepp; 00334 bool resizepp_delete; 00335 idx<T> image; 00336 double contrast; 00337 double brightness; 00338 idx<float> sizes; 00339 fstate_idx<T> finput; 00340 Tstate *input; 00341 mstate<Tstate> output; 00342 Tstate *tmp; 00343 Tstate *minput; 00344 svector<mstate<Tstate> > ppinputs; 00345 svector<mstate<Tstate> > outputs; 00346 vector<string> labels; 00347 protected: 00348 // dimensions ////////////////////////////////////////////////////////////// 00349 idxdim indim; 00350 idxdim netdim; 00351 bool netdim_fixed; 00352 // bboxes ////////////////////////////////////////////////////////////////// 00353 vector<rect<int> > original_bboxes; 00354 int bgclass; 00355 int mask_class; 00356 idx<T> mask; 00357 nms *pnms; 00358 // scales ////////////////////////////////////////////////////////////// 00359 midxdim scales; 00360 midxdim actual_scales; 00361 midxdim manual_scales; 00362 vector<double> scale_factors; 00363 uint nscales; 00364 double scales_step; 00365 double min_scale; 00366 double max_scale; 00367 t_scaling restype; 00368 // saving ////////////////////////////////////////////////////////////// 00369 bool silent; 00370 bool save_mode; 00371 string save_dir; 00372 vector<uint> save_counts; 00373 bboxes raw_bboxes; 00374 bboxes pruned_bboxes; 00375 uint min_size; 00376 uint max_size; 00377 vector<idx<T> > odetections; 00378 svector<midx<T> > ppdetections; 00379 bool bodetections; 00380 bool bppdetections; 00381 uint save_max_per_frame; 00382 bool diverse_ordering; 00383 bool mem_optimization; 00384 bool optimization_swap; 00385 bool keep_inputs; 00386 uint hzpad; 00387 uint wzpad; 00388 // printing //////////////////////////////////////////////////////////////// 00389 std::ostream &mout; 00390 std::ostream &merr; 00391 // smoothing ////////////////////////////////////////////////////////////// 00392 uint smoothing_type; 00393 idx<T> smoothing_kernel; 00394 bool initialized; 00395 string outputs_dump; 00396 bool bboxes_off; 00397 bool adapt_scales; 00398 bool scaler_mode; 00399 answer_module<T,T,T,Tstate> *answer; 00400 mstate<Tstate> answers; 00401 bool ignore_outsiders; 00402 uint corners_inference; 00403 bool corners_infered; 00404 mfidxdim itl, itr, ibl, ibr; 00405 mfidxdim pptl, pptr, ppbl, ppbr; 00406 float pre_threshold; 00407 vector<float> raw_thresholds; 00408 vector<uint> scale_indices; 00409 uint bbox_decision; 00410 mfidxdim bbox_scalings; 00411 00412 // friends ///////////////////////////////////////////////////////////////// 00413 template <typename T2, class Tstate2> friend class detector_gui; 00414 template <typename T2> friend class detection_thread; 00415 template <typename T2, class Tstate2> friend class bootstrapping; 00416 }; 00417 00418 } // end namespace ebl 00419 00420 #include "detector.hpp" 00421 00422 #endif /* DETECTOR_H_ */
1.7.4