Update of /cvsroot/pure-data/externals/grh/pix_recNN In directory sc8-pr-cvs1.sourceforge.net:/tmp/cvs-serv27874
Added Files: pix_recNN-help.pd Removed Files: help-pix_recNN.pd Log Message: new helpfile standard
--- help-pix_recNN.pd DELETED ---
--- NEW FILE: pix_recNN-help.pd --- #N canvas 871 74 498 783 10; #X obj 36 327 gemwin; #X msg 36 301 create , 1; #N canvas 75 72 765 790 pix2sig_stuff~ 0; #X obj 120 35 gemhead; #X obj 120 132 pix_texture; #X obj 119 274 outlet~; #X obj 139 185 square 4; #X obj 139 163 separator; #X obj 61 165 separator; #X obj 120 101 pix_video; #X msg 186 64 dimen 640 480; #X obj 26 36 block~ 2048; #X msg 186 38 dimen 320 240; #X msg 76 535 getprecision; #X msg 93 696 getlearnrate; #X msg 65 671 learnrate 0.2; #X msg 424 459 getneurons; #X msg 404 206 train; #X obj 31 227 inlet~; #X msg 65 647 learnrate 0.05; #X msg 381 708 getmemory; #X msg 361 639 memory 0; #X msg 361 660 memory 1; #X obj 61 252 pix_recNN; #X text 296 49 <- input dimension; #X obj 78 226 r $0-recNN; #X obj 62 564 s $0-recNN; #X msg 76 498 precision $1; #X floatatom 76 481 5 0 0 0 - - -; #X text 42 335 precision:; #X text 53 358 1: means every pixel is used in calculation; #X text 53 372 2: only every second pixel; #X text 53 386 ...; #X obj 62 411 loadbang; #X msg 407 401 neurons 2048; #X msg 407 422 neurons 64; #X obj 407 492 s $0-recNN; #X text 403 336 neurons:; #X text 416 357 nr. of neurons used in the calculation; #X text 415 370 (_MUST_ be the same as the buffersize !!!); #X text 43 615 learnrate:; #X obj 65 725 s $0-recNN; #X msg 361 681 memory 3; #X obj 361 741 s $0-recNN; #X text 343 543 memory:; #X text 356 565 this determines , how much values from the past the recurrent net considers in the calculation; #X text 357 604 (be carefull with large values !!!); #X msg 62 456 precision 1; #X msg 62 436 precision 4; #X obj 404 233 s $0-recNN; #X text 397 126 train:; #X text 417 152 trains the neural net; #X text 418 166 (the current video frame to; #X text 425 178 the current audio block); #X connect 0 0 6 0; #X connect 1 0 4 0; #X connect 1 0 5 0; #X connect 4 0 3 0; #X connect 5 0 20 0; #X connect 6 0 1 0; #X connect 7 0 6 0; #X connect 9 0 6 0; #X connect 10 0 23 0; #X connect 11 0 38 0; #X connect 12 0 38 0; #X connect 13 0 33 0; #X connect 14 0 46 0; #X connect 15 0 20 0; #X connect 16 0 38 0; #X connect 17 0 40 0; #X connect 18 0 40 0; #X connect 19 0 40 0; #X connect 20 1 2 0; #X connect 22 0 20 0; #X connect 24 0 23 0; #X connect 25 0 24 0; #X connect 30 0 45 0; #X connect 31 0 33 0; #X connect 32 0 33 0; #X connect 39 0 40 0; #X connect 44 0 23 0; #X connect 45 0 23 0; #X restore 89 542 pd pix2sig_stuff~; #X msg 110 302 0 , destroy; #X obj 116 587 unsig~; #X obj 206 432 osc~ 440; #X obj 205 456 *~; #X obj 237 456 tgl 15 0 empty empty empty 0 -6 0 8 -262144 -1 -1 0 1; #X obj 207 496 sig~ 0; #X floatatom 117 608 8 0 0 0 - - -; #X text 25 23 pix_recNN:; #X text 24 57 pix_recNN is an instument/interface. This instrument should be useful as a general experimental video interface to generate audio. You can train the neural net with playing audio samples to specific video frames in real-time. The main interest for me was not to train the net exactly to reproduce these samples , but to make experimental sounds , which are "between" all the trained samples.; #X text 22 214 (but this version is unfinished - e.g. the training algorithm must be tuned etc. - so it's only a very basic prototype...) ; #X text 207 320 <- create gemwin; #X obj 41 442 readsf~; #X obj 41 401 openpanel; #X msg 41 421 open $1; #X obj 41 380 bng 15 250 50 0 empty empty empty 0 -6 0 8 -262144 -1 -1; #X text 67 379 <- load sample for training; #X obj 122 417 tgl 25 0 empty empty empty 0 -6 0 8 -195568 -1 -1 0 1; #X floatatom 206 414 5 0 0 0 - - -; #X text 272 431 <- simple osc for training; #X text 262 497 <- to train silence; #X obj 85 463 bng 15 250 50 0 empty empty empty 0 -6 0 8 -262144 -1 -1; #X text 216 541 <- audio/video work; #X obj 90 684 dac~; #X obj 90 659 *~; #X obj 118 659 dbtorms; #X floatatom 118 641 5 0 0 0 - - -; #X text 168 638 <- outvol in dB; #X text 22 170 pix_recNN uses a 2 layer recurrent neural net (for more detailed info look at the source code.); #X text 119 737 Georg Holzmann grh@mur.at , 2004; #X connect 1 0 0 0; #X connect 2 0 4 0; #X connect 2 0 26 0; #X connect 3 0 0 0; #X connect 4 0 9 0; #X connect 5 0 6 0; #X connect 6 0 2 0; #X connect 7 0 6 1; #X connect 8 0 2 0; #X connect 14 0 2 0; #X connect 14 1 23 0; #X connect 15 0 16 0; #X connect 16 0 14 0; #X connect 17 0 15 0; #X connect 19 0 14 0; #X connect 20 0 5 0; #X connect 26 0 25 0; #X connect 26 0 25 1; #X connect 27 0 26 1; #X connect 28 0 27 0;