-
Notifications
You must be signed in to change notification settings - Fork 1
/
export_DNeuro.html
2099 lines (1959 loc) · 110 KB
/
export_DNeuro.html
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
<!DOCTYPE html>
<html class="writer-html5" lang="en" >
<head>
<meta charset="utf-8">
<meta name="viewport" content="width=device-width, initial-scale=1.0">
<title>Export: DNeuro — N2D2 documentation</title>
<link rel="stylesheet" href="_static/css/theme.css" type="text/css" />
<link rel="stylesheet" href="_static/pygments.css" type="text/css" />
<!--[if lt IE 9]>
<script src="_static/js/html5shiv.min.js"></script>
<![endif]-->
<script type="text/javascript" id="documentation_options" data-url_root="./" src="_static/documentation_options.js"></script>
<script src="_static/jquery.js"></script>
<script src="_static/underscore.js"></script>
<script src="_static/doctools.js"></script>
<script src="_static/language_data.js"></script>
<script async="async" src="https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.7/latest.js?config=TeX-AMS-MML_HTMLorMML"></script>
<script type="text/javascript" src="_static/js/theme.js"></script>
<link rel="author" title="About these documents" href="about.html" />
<link rel="index" title="Index" href="genindex.html" />
<link rel="search" title="Search" href="search.html" />
<link rel="next" title="Export: ONNX" href="export_ONNX.html" />
<link rel="prev" title="Export: TensorRT" href="export_TensorRT.html" />
</head>
<body class="wy-body-for-nav">
<div class="wy-grid-for-nav">
<nav data-toggle="wy-nav-shift" class="wy-nav-side">
<div class="wy-side-scroll">
<div class="wy-side-nav-search" >
<a href="index.html" class="icon icon-home" alt="Documentation Home"> N2D2
</a>
<div role="search">
<form id="rtd-search-form" class="wy-form" action="search.html" method="get">
<input type="text" name="q" placeholder="Search docs" />
<input type="hidden" name="check_keywords" value="yes" />
<input type="hidden" name="area" value="default" />
</form>
</div>
</div>
<div class="wy-menu wy-menu-vertical" data-spy="affix" role="navigation" aria-label="main navigation">
<p class="caption"><span class="caption-text">Introduction:</span></p>
<ul>
<li class="toctree-l1"><a class="reference internal" href="intro.html">Presentation</a></li>
<li class="toctree-l1"><a class="reference internal" href="about.html">About N2D2-IP</a></li>
<li class="toctree-l1"><a class="reference internal" href="simus.html">Performing simulations</a></li>
<li class="toctree-l1"><a class="reference internal" href="perfs_tools.html">Performance evaluation tools</a></li>
<li class="toctree-l1"><a class="reference internal" href="tuto.html">Tutorials</a></li>
</ul>
<p class="caption"><span class="caption-text">ONNX Import:</span></p>
<ul>
<li class="toctree-l1"><a class="reference internal" href="onnx_convert.html">Obtain ONNX models</a></li>
<li class="toctree-l1"><a class="reference internal" href="onnx_import.html">Import ONNX models</a></li>
<li class="toctree-l1"><a class="reference internal" href="onnx_transfer.html">Train from ONNX models</a></li>
</ul>
<p class="caption"><span class="caption-text">Quantization and Export:</span></p>
<ul class="current">
<li class="toctree-l1"><a class="reference internal" href="quant_post.html">Post-training quantization</a></li>
<li class="toctree-l1"><a class="reference internal" href="quant_qat.html">[NEW] Quantization-Aware Training</a></li>
<li class="toctree-l1"><a class="reference internal" href="export_CPP.html">Export: C++</a></li>
<li class="toctree-l1"><a class="reference internal" href="export_CPP_STM32.html">Export: C++/STM32</a></li>
<li class="toctree-l1"><a class="reference internal" href="export_TensorRT.html">Export: TensorRT</a></li>
<li class="toctree-l1 current"><a class="current reference internal" href="#">Export: DNeuro</a><ul>
<li class="toctree-l2"><a class="reference internal" href="#introduction">Introduction</a><ul>
<li class="toctree-l3"><a class="reference internal" href="#interface">Interface</a></li>
<li class="toctree-l3"><a class="reference internal" href="#supported-layers">Supported layers</a></li>
</ul>
</li>
<li class="toctree-l2"><a class="reference internal" href="#usage">Usage</a><ul>
<li class="toctree-l3"><a class="reference internal" href="#simulation">Simulation</a></li>
<li class="toctree-l3"><a class="reference internal" href="#c-emulation">C++ emulation</a></li>
<li class="toctree-l3"><a class="reference internal" href="#synthesis">Synthesis</a></li>
<li class="toctree-l3"><a class="reference internal" href="#export-parameters">Export parameters</a></li>
<li class="toctree-l3"><a class="reference internal" href="#fpga-compatibility-tables">FPGA compatibility tables</a></li>
</ul>
</li>
<li class="toctree-l2"><a class="reference internal" href="#aerial-imagery-segmentation-demo">Aerial Imagery Segmentation DEMO</a><ul>
<li class="toctree-l3"><a class="reference internal" href="#specifications">Specifications</a></li>
<li class="toctree-l3"><a class="reference internal" href="#application-preview">Application preview</a></li>
<li class="toctree-l3"><a class="reference internal" href="#dneuro-generation">DNeuro generation</a></li>
</ul>
</li>
<li class="toctree-l2"><a class="reference internal" href="#face-detection-demo">Face Detection DEMO</a></li>
</ul>
</li>
<li class="toctree-l1"><a class="reference internal" href="export_ONNX.html">Export: ONNX</a></li>
<li class="toctree-l1"><a class="reference internal" href="export_legacy.html">Export: other / legacy</a></li>
</ul>
<p class="caption"><span class="caption-text">INI File Interface:</span></p>
<ul>
<li class="toctree-l1"><a class="reference internal" href="ini_intro.html">Introduction</a></li>
<li class="toctree-l1"><a class="reference internal" href="ini_databases.html">Databases</a></li>
<li class="toctree-l1"><a class="reference internal" href="ini_data_analysis.html">Stimuli data analysis</a></li>
<li class="toctree-l1"><a class="reference internal" href="ini_environment.html">Stimuli provider (Environment)</a></li>
<li class="toctree-l1"><a class="reference internal" href="ini_layers.html">Network Layers</a></li>
<li class="toctree-l1"><a class="reference internal" href="ini_target.html">Targets (outputs & losses)</a></li>
<li class="toctree-l1"><a class="reference internal" href="adversarial.html">Adversarial module</a></li>
</ul>
<p class="caption"><span class="caption-text">Python API:</span></p>
<ul>
<li class="toctree-l1"><a class="reference internal" href="containers.html">Containers</a></li>
<li class="toctree-l1"><a class="reference internal" href="cells.html">Cells</a></li>
<li class="toctree-l1"><a class="reference internal" href="databases.html">Databases</a></li>
<li class="toctree-l1"><a class="reference internal" href="stimuliprovider.html">StimuliProvider</a></li>
<li class="toctree-l1"><a class="reference internal" href="deepnet.html">DeepNet</a></li>
</ul>
<p class="caption"><span class="caption-text">C++ API / Developer:</span></p>
<ul>
<li class="toctree-l1"><a class="reference internal" href="dev_intro.html">Introduction</a></li>
</ul>
</div>
</div>
</nav>
<section data-toggle="wy-nav-shift" class="wy-nav-content-wrap">
<nav class="wy-nav-top" aria-label="top navigation">
<i data-toggle="wy-nav-top" class="fa fa-bars"></i>
<a href="index.html">N2D2</a>
</nav>
<div class="wy-nav-content">
<div class="rst-content">
<div role="navigation" aria-label="breadcrumbs navigation">
<ul class="wy-breadcrumbs">
<li><a href="index.html" class="icon icon-home"></a> »</li>
<li>Export: DNeuro</li>
<li class="wy-breadcrumbs-aside">
<a href="_sources/export_DNeuro.rst.txt" rel="nofollow"> View page source</a>
</li>
</ul>
<hr/>
</div>
<div role="main" class="document" itemscope="itemscope" itemtype="http://schema.org/Article">
<div itemprop="articleBody">
<div class="section" id="export-dneuro">
<h1>Export: DNeuro<a class="headerlink" href="#export-dneuro" title="Permalink to this headline">¶</a></h1>
<p><strong>N2D2-IP only: available upon request.</strong></p>
<dl class="simple">
<dt>Export type: <code class="docutils literal notranslate"><span class="pre">DNeuro_V2</span></code></dt><dd><p>DNeuro RTL export for FPGA.</p>
</dd>
</dl>
<div class="highlight-default notranslate"><div class="highlight"><pre><span></span><span class="n">n2d2</span> <span class="n">MobileNet_ONNX</span><span class="o">.</span><span class="n">ini</span> <span class="o">-</span><span class="n">seed</span> <span class="mi">1</span> <span class="o">-</span><span class="n">w</span> <span class="o">/</span><span class="n">dev</span><span class="o">/</span><span class="n">null</span> <span class="o">-</span><span class="n">export</span> <span class="n">DNeuro_V2</span>
</pre></div>
</div>
<div class="section" id="introduction">
<h2>Introduction<a class="headerlink" href="#introduction" title="Permalink to this headline">¶</a></h2>
<p>DNeuro is a synthetizable dataflow architecture, optimized for deep
convolutional neural networks (CNN). It allows a fine grain allocation
control of the DSP and memory resources, for each layer in a network.
Globally, the FPGA resource usage can be maximized for a given network
topology in order to minimize its latency.</p>
<p>The main features of the DNeuro are:</p>
<ul class="simple">
<li><p>Data flow architecture requiring few memory (potentially <strong>no DDR</strong>);</p></li>
<li><p>Very high use rate of the DSP per cycle (> 90%);</p></li>
<li><p>Configurable precision (integers from 2 to 16 bits, typically 8 bits);</p></li>
<li><p>Up to 4 MAC/DSP operations per cycle.</p></li>
</ul>
<p>The DNeuro is composed of specialized computing blocs, corresponding to
specific type and configuration of layers (convolution, max pooling…),
that can be chained to form a full neural network. The bloc allocation
and chaining is done automatically with N2D2.</p>
<div class="section" id="interface">
<h3>Interface<a class="headerlink" href="#interface" title="Permalink to this headline">¶</a></h3>
<p>The DNeuro interface is extremely simple and behaves like a
pipeline/FIFO.</p>
<p>An example of the top-level DNeuro RTL entity is described below, for
one input channel and 3 output channels:</p>
<div class="highlight-vhdl notranslate"><div class="highlight"><pre><span></span><span class="c1">-- Input size: 1*640*480</span>
<span class="c1">-- Output size: 3*80*60</span>
<span class="k">entity</span> <span class="nc">network</span> <span class="k">is</span>
<span class="k">generic</span> <span class="p">(</span>
<span class="k">constant</span> <span class="n">G_BATCH_SIZE</span><span class="o">:</span> <span class="kt">positive</span> <span class="o">:=</span> <span class="mi">1</span><span class="p">;</span>
<span class="k">constant</span> <span class="n">G_FIFO_DEPTH</span><span class="o">:</span> <span class="kt">positive</span> <span class="o">:=</span> <span class="mi">1</span><span class="p">;</span>
<span class="k">constant</span> <span class="n">G_DATA_LENGTH</span><span class="o">:</span> <span class="kt">positive</span> <span class="o">:=</span> <span class="mi">8</span><span class="p">;</span>
<span class="k">constant</span> <span class="n">G_ACC_S_LENGTH</span><span class="o">:</span> <span class="kt">positive</span> <span class="o">:=</span> <span class="mi">18</span><span class="p">;</span>
<span class="k">constant</span> <span class="n">G_NB_OUTPUTS_INST_N_1_ENV</span><span class="o">:</span> <span class="kt">positive</span> <span class="o">:=</span> <span class="mi">1</span><span class="p">;</span>
<span class="k">constant</span> <span class="n">G_NB_OUTPUTS_MERG_N_1_ENV</span><span class="o">:</span> <span class="kt">positive</span> <span class="o">:=</span> <span class="mi">1</span>
<span class="p">);</span>
<span class="k">port</span> <span class="p">(</span>
<span class="n">clk</span> <span class="o">:</span> <span class="k">in</span> <span class="kt">std_logic</span><span class="p">;</span>
<span class="n">rstn</span> <span class="o">:</span> <span class="k">in</span> <span class="kt">std_logic</span><span class="p">;</span>
<span class="n">i_data</span> <span class="o">:</span> <span class="k">in</span> <span class="kt">std_logic_vector</span> <span class="p">((</span><span class="n">G_DATA_LENGTH</span><span class="o">*</span><span class="n">G_BATCH_SIZE</span><span class="p">)</span><span class="o">-</span><span class="mi">1</span> <span class="k">downto</span> <span class="mi">0</span><span class="p">);</span>
<span class="n">i_valid_data</span> <span class="o">:</span> <span class="k">in</span> <span class="kt">std_logic</span><span class="p">;</span>
<span class="n">o_en</span> <span class="o">:</span> <span class="k">out</span> <span class="kt">std_logic</span><span class="p">;</span>
<span class="n">o_data</span> <span class="o">:</span> <span class="k">out</span> <span class="kt">std_logic_vector</span> <span class="p">((</span><span class="mi">3</span><span class="o">*</span><span class="n">G_DATA_LENGTH</span><span class="o">*</span><span class="n">G_BATCH_SIZE</span><span class="p">)</span><span class="o">-</span><span class="mi">1</span> <span class="k">downto</span> <span class="mi">0</span><span class="p">);</span>
<span class="n">o_valid_data</span> <span class="o">:</span> <span class="k">out</span> <span class="kt">std_logic</span><span class="p">;</span>
<span class="n">i_en</span> <span class="o">:</span> <span class="k">in</span> <span class="kt">std_logic</span>
<span class="p">);</span>
<span class="k">end</span> <span class="nc">network</span><span class="p">;</span>
</pre></div>
</div>
</div>
<div class="section" id="supported-layers">
<h3>Supported layers<a class="headerlink" href="#supported-layers" title="Permalink to this headline">¶</a></h3>
<table class="docutils align-default">
<colgroup>
<col style="width: 49%" />
<col style="width: 8%" />
<col style="width: 42%" />
</colgroup>
<thead>
<tr class="row-odd"><th class="head"><p>Layer type</p></th>
<th class="head"><p>Support</p></th>
<th class="head"><p>Comments</p></th>
</tr>
</thead>
<tbody>
<tr class="row-even"><td><p>Dropout</p></td>
<td><p>n.a.</p></td>
<td><p>removed during export</p></td>
</tr>
<tr class="row-odd"><td><p>Fc</p></td>
<td><p><span class="raw-html"><font color="green"></span> ✓ <span class="raw-html"></font></span></p></td>
<td><p>implemented with Conv during export</p></td>
</tr>
<tr class="row-even"><td colspan="3"><p><em>InnerProduct</em> <span class="math notranslate nohighlight">\(\rightarrow\)</span> see Fc</p></td>
</tr>
<tr class="row-odd"><td><p>Transformation</p></td>
<td><p>✗</p></td>
<td></td>
</tr>
<tr class="row-even"><td><p>BatchNorm</p></td>
<td><p>n.a.</p></td>
<td><p>merged with Conv during export with <code class="docutils literal notranslate"><span class="pre">-fuse</span></code> option</p></td>
</tr>
<tr class="row-odd"><td><p>Conv</p></td>
<td><p><span class="raw-html"><font color="green"></span> ✓ <span class="raw-html"></font></span></p></td>
<td></td>
</tr>
<tr class="row-even"><td colspan="3"><p><em>Concat</em> <span class="math notranslate nohighlight">\(\rightarrow\)</span> implicit for Conv/Deconv/Pool/Fc</p></td>
</tr>
<tr class="row-odd"><td><p>Deconv</p></td>
<td><p>✗</p></td>
<td></td>
</tr>
<tr class="row-even"><td><p>ElemWise</p></td>
<td><p><span class="raw-html"><font color="green"></span> ✓ <span class="raw-html"></font></span></p></td>
<td><p><em>Sum</em> operation only</p></td>
</tr>
<tr class="row-odd"><td colspan="3"><p><em>EltWise</em> <span class="math notranslate nohighlight">\(\rightarrow\)</span> see ElemWise</p></td>
</tr>
<tr class="row-even"><td colspan="3"><p><em>Flatten</em> <span class="math notranslate nohighlight">\(\rightarrow\)</span> implicit to Fc/Rbf</p></td>
</tr>
<tr class="row-odd"><td><p>LRN</p></td>
<td><p>✗</p></td>
<td></td>
</tr>
<tr class="row-even"><td colspan="3"><p><em>Maxout</em> <span class="math notranslate nohighlight">\(\rightarrow\)</span> see Pool</p></td>
</tr>
<tr class="row-odd"><td><p>Padding</p></td>
<td><p><span class="raw-html"><font color="green"></span> ✓ <span class="raw-html"></font></span></p></td>
<td><p>merged with Conv/Pool during export</p></td>
</tr>
<tr class="row-even"><td><p>Pool</p></td>
<td><p><span class="raw-html"><font color="green"></span> ✓ <span class="raw-html"></font></span></p></td>
<td><p><em>Max</em> operation only</p></td>
</tr>
<tr class="row-odd"><td><p>Resize</p></td>
<td><p><span class="raw-html"><font color="green"></span> ✓ <span class="raw-html"></font></span></p></td>
<td><p><em>NearestNeighbor</em> mode only</p></td>
</tr>
<tr class="row-even"><td><p>Softmax</p></td>
<td><p>✗</p></td>
<td></td>
</tr>
<tr class="row-odd"><td colspan="3"><p><em>SortLabel</em> <span class="math notranslate nohighlight">\(\rightarrow\)</span> see .Target*</p></td>
</tr>
<tr class="row-even"><td><p>Unpool</p></td>
<td><p>✗</p></td>
<td></td>
</tr>
<tr class="row-odd"><td colspan="3"><p><em>Upscale</em> <span class="math notranslate nohighlight">\(\rightarrow\)</span> see Resize</p></td>
</tr>
<tr class="row-even"><td><p>.Target*</p></td>
<td><p><span class="raw-html"><font color="green"></span> ✓ <span class="raw-html"></font></span></p></td>
<td><p>top-1 sorting</p></td>
</tr>
</tbody>
</table>
<table class="docutils align-default">
<colgroup>
<col style="width: 43%" />
<col style="width: 11%" />
<col style="width: 46%" />
</colgroup>
<thead>
<tr class="row-odd"><th class="head"><p>Activation type</p></th>
<th class="head"><p>Support</p></th>
<th class="head"><p>Specificities</p></th>
</tr>
</thead>
<tbody>
<tr class="row-even"><td><p>Linear</p></td>
<td><p><span class="raw-html"><font color="green"></span> ✓ <span class="raw-html"></font></span></p></td>
<td><p>saturated arithmetic</p></td>
</tr>
<tr class="row-odd"><td><p>Logistic</p></td>
<td><p><span class="raw-html"><font color="green"></span> ✓ <span class="raw-html"></font></span></p></td>
<td><p>saturation approximation, configurable zero,
up to two configurable thresholds</p></td>
</tr>
<tr class="row-even"><td colspan="3"><p><em>ReLU</em> <span class="math notranslate nohighlight">\(\rightarrow\)</span> see Rectifier</p></td>
</tr>
<tr class="row-odd"><td colspan="3"><p><em>bReLU</em> <span class="math notranslate nohighlight">\(\rightarrow\)</span> see Rectifier</p></td>
</tr>
<tr class="row-even"><td><p>Rectifier</p></td>
<td><p><span class="raw-html"><font color="green"></span> ✓ <span class="raw-html"></font></span></p></td>
<td><p>saturated arithmetic (positive values)</p></td>
</tr>
<tr class="row-odd"><td><p>Saturation</p></td>
<td><p><span class="raw-html"><font color="green"></span> ✓ <span class="raw-html"></font></span></p></td>
<td></td>
</tr>
<tr class="row-even"><td><p>Softplus</p></td>
<td><p>✗</p></td>
<td></td>
</tr>
<tr class="row-odd"><td><p>Tanh</p></td>
<td><p>✗</p></td>
<td></td>
</tr>
</tbody>
</table>
</div>
</div>
<div class="section" id="usage">
<h2>Usage<a class="headerlink" href="#usage" title="Permalink to this headline">¶</a></h2>
<div class="section" id="simulation">
<h3>Simulation<a class="headerlink" href="#simulation" title="Permalink to this headline">¶</a></h3>
<p>When a network is exported, test vectors are exported automatically too, if
the <code class="docutils literal notranslate"><span class="pre">-db-export</span></code> command line option value is > 0 (by default, the full test
set is exported). All the test vectors are exported for the C++ emulator, while
only the first image is pre-loaded as a test vector for the RTL simulation in
the <code class="docutils literal notranslate"><span class="pre">RTL/NETWORK/TB/network_tb.vhd</span></code> file. This testbench is configured with a
clock frequency of 100MHz (regardless of the <code class="docutils literal notranslate"><span class="pre">EstimationFrequency</span></code> export
parameter). The testbench reads 3 times the same (first) image and outputs the
results in the <code class="docutils literal notranslate"><span class="pre">out_file/out.txt</span></code> file, located in <code class="docutils literal notranslate"><span class="pre">RTL/NETWORK/simu/VsimTOOL</span></code>
for ModelSim.</p>
<div class="highlight-default notranslate"><div class="highlight"><pre><span></span><span class="n">cd</span> <span class="n">RTL</span><span class="o">/</span><span class="n">NETWORK</span><span class="o">/</span><span class="n">simu</span>
<span class="n">make</span> <span class="n">vsim</span>
</pre></div>
</div>
</div>
<div class="section" id="c-emulation">
<h3>C++ emulation<a class="headerlink" href="#c-emulation" title="Permalink to this headline">¶</a></h3>
<p>The DNeuro export comes with a C++ bit-accurate emulator.</p>
<p>By default, the provided emulator will use the same parameters as the ones
defined in the export. For testing purposes it is possible to change the
accumulation size by defining the <code class="docutils literal notranslate"><span class="pre">ACC_NB_BITS</span></code> variable.</p>
<div class="highlight-default notranslate"><div class="highlight"><pre><span></span><span class="n">cd</span> <span class="n">EMULATOR</span>
<span class="n">CXXFLAGS</span><span class="o">=</span><span class="s2">"-DACC_NB_BITS=18"</span> <span class="n">make</span>
<span class="o">./</span><span class="n">dneuro_v2_emulator</span>
</pre></div>
</div>
<p>When running the emulator, all the exported images are evaluated by default, and
a global score is computed from individual images good or bad classifications.
It is possible to evaluate a single image with the following command line
argument:</p>
<div class="highlight-default notranslate"><div class="highlight"><pre><span></span><span class="o">./</span><span class="n">dneuro_v2_emulator</span> <span class="o">-</span><span class="n">stimulus</span> <span class="n">stimuli</span><span class="o">/</span><span class="n">env00</span><span class="o">.</span><span class="n">ppm</span>
</pre></div>
</div>
<p>The <code class="docutils literal notranslate"><span class="pre">stimuli/env00.ppm</span></code> is an already pre-processed image automatically
exported by N2D2 and ready to be feed at the input of the neural network. The
emulator generates for each network’s layer an output file
<em>layer_name_output.txt</em> containing the output tensor values of the layer, as
expected for the DNeuro IP.</p>
</div>
<div class="section" id="synthesis">
<h3>Synthesis<a class="headerlink" href="#synthesis" title="Permalink to this headline">¶</a></h3>
<p>To generate a project ready for synthesis in Vivado or Quartus, use the scripts
provided in <code class="docutils literal notranslate"><span class="pre">RTL/NETWORK/simu/PythonTOOL</span></code>. To generate a Vivado project, run:</p>
<div class="highlight-default notranslate"><div class="highlight"><pre><span></span><span class="n">cd</span> <span class="n">RTL</span><span class="o">/</span><span class="n">NETWORK</span><span class="o">/</span><span class="n">simu</span>
<span class="n">python</span> <span class="n">PythonTOOL</span><span class="o">/</span><span class="n">vivadoGenerate</span><span class="o">.</span><span class="n">py</span>
</pre></div>
</div>
<p>This script creates a new project in <code class="docutils literal notranslate"><span class="pre">RTL/NETWORK/simu/VivadoTOOL/project_export_DNeuro</span></code>.
Do not forget to change the default project’s part.</p>
<div class="admonition warning">
<p class="admonition-title">Warning</p>
<p>Do not create a project and add the sources manually, as the sources
organization into libraries will not be setup properly: the sources in the
directories <code class="docutils literal notranslate"><span class="pre">CONV_COMMON</span></code>, <code class="docutils literal notranslate"><span class="pre">CONV_Tn_Oy_CHy_K1_Sy_P1</span></code> and
<code class="docutils literal notranslate"><span class="pre">CONV_Tn_Oy_CHy_K1_Sy_Pn</span></code> must be placed in libraries of the same name!</p>
</div>
</div>
<div class="section" id="export-parameters">
<h3>Export parameters<a class="headerlink" href="#export-parameters" title="Permalink to this headline">¶</a></h3>
<p>Extra parameters can be passed during export using the
<code class="docutils literal notranslate"><span class="pre">-export-parameters</span> <span class="pre">params.ini</span></code> command line argument. The parameters must be
saved in an INI-like file.</p>
<p>List of general available parameters:</p>
<table class="docutils align-default">
<colgroup>
<col style="width: 35%" />
<col style="width: 65%" />
</colgroup>
<thead>
<tr class="row-odd"><th class="head"><p>Argument [default value]</p></th>
<th class="head"><p>Description</p></th>
</tr>
</thead>
<tbody>
<tr class="row-even"><td><p><code class="docutils literal notranslate"><span class="pre">NbDSPs</span></code></p></td>
<td><p>Set the maximum number of DSPs that the network can use on the FPGA</p></td>
</tr>
<tr class="row-odd"><td><p><code class="docutils literal notranslate"><span class="pre">NbMemoryBytes</span></code></p></td>
<td><p>Set the maximum memory, in bytes, that the network can use on the FPGA</p></td>
</tr>
<tr class="row-even"><td><p><code class="docutils literal notranslate"><span class="pre">Network</span></code> [network]</p></td>
<td><p>Name of the top-level HDL entity</p></td>
</tr>
<tr class="row-odd"><td><p><code class="docutils literal notranslate"><span class="pre">EstimationFrequency</span></code> [200]</p></td>
<td><p>Frequency used for the FPS estimation given by the export</p></td>
</tr>
<tr class="row-even"><td><p><code class="docutils literal notranslate"><span class="pre">AccumulationNbBits</span></code> [2.DATA_LENGTH+4]</p></td>
<td><p>Number of bits to use for the accumulation</p></td>
</tr>
</tbody>
</table>
<p>Output map class conversion to RGB settings:</p>
<table class="docutils align-default">
<colgroup>
<col style="width: 35%" />
<col style="width: 65%" />
</colgroup>
<thead>
<tr class="row-odd"><th class="head"><p>Argument [default value]</p></th>
<th class="head"><p>Description</p></th>
</tr>
</thead>
<tbody>
<tr class="row-even"><td><p><code class="docutils literal notranslate"><span class="pre">OutputMapToRGB</span></code> [0]</p></td>
<td><p>If true (1), add an extra layer at the end of the network that converts the output of the network to an RGB output</p></td>
</tr>
<tr class="row-odd"><td><p><code class="docutils literal notranslate"><span class="pre">OutputMapToRGBBackgroundClass</span></code> []</p></td>
<td><p>When <code class="docutils literal notranslate"><span class="pre">OutputMapToRGB</span></code> is 1, set the class that is used for background objects. The overlay color for this class will
be transparent</p></td>
</tr>
<tr class="row-even"><td><p><code class="docutils literal notranslate"><span class="pre">OutputMapToRGBColorMasks</span></code> []</p></td>
<td><p>When <code class="docutils literal notranslate"><span class="pre">OutputMapToRGB</span></code> is 1, list of colors to use for the classes</p></td>
</tr>
<tr class="row-odd"><td><p><code class="docutils literal notranslate"><span class="pre">OutputMapToRGBBinaryThresholdUpper</span></code> [0]</p></td>
<td><p>Upper threshold for binary outputs</p></td>
</tr>
<tr class="row-even"><td><p><code class="docutils literal notranslate"><span class="pre">OutputMapToRGBBinaryThresholdLower</span></code> [0]</p></td>
<td><p>Lower threshold for binary outputs</p></td>
</tr>
</tbody>
</table>
<p>Internal per layer settings (for debug purpose only!):</p>
<table class="docutils align-default">
<colgroup>
<col style="width: 35%" />
<col style="width: 65%" />
</colgroup>
<thead>
<tr class="row-odd"><th class="head"><p>Argument [default value]</p></th>
<th class="head"><p>Description</p></th>
</tr>
</thead>
<tbody>
<tr class="row-even"><td><p><code class="docutils literal notranslate"><span class="pre">RTLType</span></code> []</p></td>
<td><p>Specific name of the RTL library module to use for this layer</p></td>
</tr>
<tr class="row-odd"><td><p><code class="docutils literal notranslate"><span class="pre">NbChannelsInstantiation</span></code> []</p></td>
<td><p>Specific number of channels to instantiate</p></td>
</tr>
<tr class="row-even"><td><p><code class="docutils literal notranslate"><span class="pre">NbOutputsInstantiation</span></code> []</p></td>
<td><p>Specific number of outputs to instantiate</p></td>
</tr>
<tr class="row-odd"><td><p><code class="docutils literal notranslate"><span class="pre">KernelHeightInstantiation</span></code> []</p></td>
<td><p>Specific number of kernel height to instantiate</p></td>
</tr>
<tr class="row-even"><td><p><code class="docutils literal notranslate"><span class="pre">KernelWidthInstantiation</span></code> []</p></td>
<td><p>Specific number of kernel width to instantiate</p></td>
</tr>
</tbody>
</table>
</div>
<div class="section" id="fpga-compatibility-tables">
<h3>FPGA compatibility tables<a class="headerlink" href="#fpga-compatibility-tables" title="Permalink to this headline">¶</a></h3>
<dl>
<dt>Legend:</dt><dd><div class="line-block">
<div class="line"><span class="math notranslate nohighlight">\(\bullet\)</span> should be OK for the standard 224x224 input, but depends on the resolution;</div>
<div class="line"><span class="raw-html"><font color="blue"></span> • <span class="raw-html"></font></span> should be OK for the standard 224x224 input using also the UltraRAM, but depends on the resolution (Xilinx FPGA only);</div>
<div class="line"><span class="raw-html"><font color="silver"></span> • <span class="raw-html"></font></span> M20K memory may be insufficient depending on the resolution;</div>
<div class="line"><span class="raw-html"><font color="orange"></span> • <span class="raw-html"></font></span> there is a better equivalent neural network (see on the same column);</div>
<div class="line"><span class="raw-html"><font color="orange"></span> ◦ <span class="raw-html"></font></span> using an alternative neural network is possible with a small accuracy loss.</div>
</div>
</dd>
<dt>Arria 10</dt><dd><p>Neural networks compatibility table with DNeuro, in terms of memory requirement.</p>
</dd>
</dl>
<table class="docutils align-default">
<colgroup>
<col style="width: 14%" />
<col style="width: 8%" />
<col style="width: 8%" />
<col style="width: 11%" />
<col style="width: 11%" />
<col style="width: 8%" />
<col style="width: 8%" />
<col style="width: 11%" />
<col style="width: 11%" />
<col style="width: 11%" />
</colgroup>
<thead>
<tr class="row-odd"><th class="head" rowspan="2"><p><strong>Arria 10</strong></p></th>
<th class="head"><p><strong>GX/SX</strong></p></th>
<th class="head"><p><strong>GX/SX</strong></p></th>
<th class="head"><p><strong>GX/SX</strong></p></th>
<th class="head"><p><strong>GX/SX</strong></p></th>
<th class="head"><p><strong>GX/SX</strong></p></th>
<th class="head"><p><strong>GX/SX</strong></p></th>
<th class="head"><p><strong>GX/SX</strong></p></th>
<th class="head"><p><strong>GX</strong></p></th>
<th class="head"><p><strong>GX</strong></p></th>
</tr>
<tr class="row-even"><th class="head"><p><strong>160</strong></p></th>
<th class="head"><p><strong>220</strong></p></th>
<th class="head"><p><strong>270</strong></p></th>
<th class="head"><p><strong>320</strong></p></th>
<th class="head"><p><strong>480</strong></p></th>
<th class="head"><p><strong>570</strong></p></th>
<th class="head"><p><strong>660</strong></p></th>
<th class="head"><p><strong>900</strong></p></th>
<th class="head"><p><strong>1150</strong></p></th>
</tr>
<tr class="row-odd"><th class="head"><p><strong>M20K (MB)</strong></p></th>
<th class="head"><p>1.12</p></th>
<th class="head"><p>1.37</p></th>
<th class="head"><p>1.87</p></th>
<th class="head"><p>2.12</p></th>
<th class="head"><p>3.5</p></th>
<th class="head"><p>4.37</p></th>
<th class="head"><p>5.25</p></th>
<th class="head"><p>5.87</p></th>
<th class="head"><p>6.62</p></th>
</tr>
<tr class="row-even"><th class="head"><p><strong>DSP</strong></p></th>
<th class="head"><p>156</p></th>
<th class="head"><p>191</p></th>
<th class="head"><p>830</p></th>
<th class="head"><p>985</p></th>
<th class="head"><p>1,368</p></th>
<th class="head"><p>1,523</p></th>
<th class="head"><p>1,688</p></th>
<th class="head"><p>1,518</p></th>
<th class="head"><p>1,518</p></th>
</tr>
<tr class="row-odd"><th class="head"><p><strong>Mult. (MAC/c.)</strong></p></th>
<th class="head"><p>312</p></th>
<th class="head"><p>382</p></th>
<th class="head"><p>1,660</p></th>
<th class="head"><p>1,970</p></th>
<th class="head"><p>2,736</p></th>
<th class="head"><p>3,046</p></th>
<th class="head"><p>3,376</p></th>
<th class="head"><p>3,036</p></th>
<th class="head"><p>3,036</p></th>
</tr>
</thead>
<tbody>
<tr class="row-even"><td><p>MobileNet_v1_0.25</p></td>
<td><p><span class="math notranslate nohighlight">\(\bullet\)</span></p></td>
<td><p><span class="math notranslate nohighlight">\(\bullet\)</span></p></td>
<td><p><span class="math notranslate nohighlight">\(\bullet\)</span></p></td>
<td><p><span class="math notranslate nohighlight">\(\bullet\)</span></p></td>
<td><p><span class="math notranslate nohighlight">\(\bullet\)</span></p></td>
<td><p><span class="math notranslate nohighlight">\(\bullet\)</span></p></td>
<td><p><span class="math notranslate nohighlight">\(\bullet\)</span></p></td>
<td><p><span class="math notranslate nohighlight">\(\bullet\)</span></p></td>
<td><p><span class="math notranslate nohighlight">\(\bullet\)</span></p></td>
</tr>
<tr class="row-odd"><td><p>MobileNet_v1_0.5</p></td>
<td></td>
<td></td>
<td><p><span class="raw-html"><font color="silver"></span> • <span class="raw-html"></font></span></p></td>
<td><p><span class="raw-html"><font color="silver"></span> • <span class="raw-html"></font></span></p></td>
<td><p><span class="math notranslate nohighlight">\(\bullet\)</span></p></td>
<td><p><span class="math notranslate nohighlight">\(\bullet\)</span></p></td>
<td><p><span class="math notranslate nohighlight">\(\bullet\)</span></p></td>
<td><p><span class="math notranslate nohighlight">\(\bullet\)</span></p></td>
<td><p><span class="math notranslate nohighlight">\(\bullet\)</span></p></td>
</tr>
<tr class="row-even"><td><p>MobileNet_v1_0.75</p></td>
<td></td>
<td></td>
<td></td>
<td></td>
<td><p><span class="raw-html"><font color="silver"></span> • <span class="raw-html"></font></span></p></td>
<td><p><span class="raw-html"><font color="silver"></span> • <span class="raw-html"></font></span></p></td>
<td><p><span class="math notranslate nohighlight">\(\bullet\)</span></p></td>
<td><p><span class="math notranslate nohighlight">\(\bullet\)</span></p></td>
<td><p><span class="math notranslate nohighlight">\(\bullet\)</span></p></td>
</tr>
<tr class="row-odd"><td><p>MobileNet_v1_1.0</p></td>
<td></td>
<td></td>
<td></td>
<td></td>
<td></td>
<td></td>
<td><p><span class="raw-html"><font color="silver"></span> • <span class="raw-html"></font></span></p></td>
<td><p><span class="raw-html"><font color="silver"></span> • <span class="raw-html"></font></span></p></td>
<td><p><span class="raw-html"><font color="silver"></span> • <span class="raw-html"></font></span></p></td>
</tr>
<tr class="row-even"><td><p>SqueezeNet_v1.0</p></td>
<td></td>
<td></td>
<td><p><span class="raw-html"><font color="orange"></span> • <span class="raw-html"></font></span> <span class="raw-html"><font color="silver"></span> • <span class="raw-html"></font></span></p></td>
<td><p><span class="raw-html"><font color="orange"></span> • <span class="raw-html"></font></span> <span class="raw-html"><font color="silver"></span> • <span class="raw-html"></font></span></p></td>
<td><p><span class="math notranslate nohighlight">\(\bullet\)</span></p></td>
<td><p><span class="math notranslate nohighlight">\(\bullet\)</span></p></td>
<td><p><span class="math notranslate nohighlight">\(\bullet\)</span></p></td>
<td><p><span class="math notranslate nohighlight">\(\bullet\)</span></p></td>
<td><p><span class="math notranslate nohighlight">\(\bullet\)</span></p></td>
</tr>
<tr class="row-odd"><td><p>SqueezeNet_v1.1</p></td>
<td></td>
<td></td>
<td><p><span class="raw-html"><font color="orange"></span> • <span class="raw-html"></font></span> <span class="raw-html"><font color="silver"></span> • <span class="raw-html"></font></span></p></td>
<td><p><span class="raw-html"><font color="orange"></span> • <span class="raw-html"></font></span> <span class="raw-html"><font color="silver"></span> • <span class="raw-html"></font></span></p></td>
<td><p><span class="math notranslate nohighlight">\(\bullet\)</span></p></td>
<td><p><span class="math notranslate nohighlight">\(\bullet\)</span></p></td>
<td><p><span class="math notranslate nohighlight">\(\bullet\)</span></p></td>
<td><p><span class="math notranslate nohighlight">\(\bullet\)</span></p></td>
<td><p><span class="math notranslate nohighlight">\(\bullet\)</span></p></td>
</tr>
<tr class="row-even"><td><p>MobileNet_v2_0.35</p></td>
<td></td>
<td></td>
<td></td>
<td><p><span class="raw-html"><font color="silver"></span> • <span class="raw-html"></font></span></p></td>
<td><p><span class="math notranslate nohighlight">\(\bullet\)</span></p></td>
<td><p><span class="math notranslate nohighlight">\(\bullet\)</span></p></td>
<td><p><span class="math notranslate nohighlight">\(\bullet\)</span></p></td>
<td><p><span class="math notranslate nohighlight">\(\bullet\)</span></p></td>
<td><p><span class="math notranslate nohighlight">\(\bullet\)</span></p></td>
</tr>
<tr class="row-odd"><td><p>MobileNet_v2_0.5</p></td>
<td></td>
<td></td>
<td></td>
<td></td>
<td><p><span class="raw-html"><font color="silver"></span> • <span class="raw-html"></font></span></p></td>
<td><p><span class="math notranslate nohighlight">\(\bullet\)</span></p></td>
<td><p><span class="math notranslate nohighlight">\(\bullet\)</span></p></td>
<td><p><span class="math notranslate nohighlight">\(\bullet\)</span></p></td>
<td><p><span class="math notranslate nohighlight">\(\bullet\)</span></p></td>
</tr>
<tr class="row-even"><td><p>MobileNet_v2_0.75</p></td>
<td></td>
<td></td>
<td></td>
<td></td>
<td><p><span class="raw-html"><font color="silver"></span> • <span class="raw-html"></font></span></p></td>
<td><p><span class="raw-html"><font color="silver"></span> • <span class="raw-html"></font></span></p></td>
<td><p><span class="math notranslate nohighlight">\(\bullet\)</span></p></td>
<td><p><span class="math notranslate nohighlight">\(\bullet\)</span></p></td>
<td><p><span class="math notranslate nohighlight">\(\bullet\)</span></p></td>
</tr>
<tr class="row-odd"><td><p>MobileNet_v2_1.0</p></td>
<td></td>
<td></td>
<td></td>
<td></td>
<td></td>
<td><p><span class="raw-html"><font color="silver"></span> • <span class="raw-html"></font></span></p></td>
<td><p><span class="raw-html"><font color="silver"></span> • <span class="raw-html"></font></span></p></td>
<td><p><span class="raw-html"><font color="silver"></span> • <span class="raw-html"></font></span></p></td>
<td><p><span class="raw-html"><font color="silver"></span> • <span class="raw-html"></font></span></p></td>
</tr>
<tr class="row-even"><td><p>MobileNet_v2_1.3</p></td>
<td></td>
<td></td>
<td></td>
<td></td>
<td></td>
<td></td>
<td></td>
<td><p><span class="raw-html"><font color="silver"></span> • <span class="raw-html"></font></span></p></td>
<td><p><span class="raw-html"><font color="silver"></span> • <span class="raw-html"></font></span></p></td>
</tr>
<tr class="row-odd"><td><p>MobileNet_v2_1.4</p></td>
<td></td>
<td></td>
<td></td>
<td></td>
<td></td>
<td></td>
<td></td>
<td></td>
<td><p><span class="raw-html"><font color="silver"></span> • <span class="raw-html"></font></span></p></td>
</tr>
<tr class="row-even"><td><p>AlexNet</p></td>
<td></td>
<td></td>
<td><p><span class="raw-html"><font color="orange"></span> • <span class="raw-html"></font></span></p></td>
<td><p><span class="raw-html"><font color="orange"></span> • <span class="raw-html"></font></span></p></td>
<td><p><span class="raw-html"><font color="orange"></span> • <span class="raw-html"></font></span></p></td>
<td><p><span class="raw-html"><font color="orange"></span> • <span class="raw-html"></font></span></p></td>
<td><p><span class="raw-html"><font color="orange"></span> • <span class="raw-html"></font></span></p></td>
<td><p><span class="raw-html"><font color="orange"></span> • <span class="raw-html"></font></span></p></td>
<td><p><span class="raw-html"><font color="orange"></span> • <span class="raw-html"></font></span></p></td>
</tr>
<tr class="row-odd"><td><p>VGG-16</p></td>
<td></td>
<td></td>
<td></td>
<td></td>
<td></td>
<td><p><span class="raw-html"><font color="orange"></span> • <span class="raw-html"></font></span></p></td>
<td><p><span class="raw-html"><font color="orange"></span> • <span class="raw-html"></font></span> <span class="raw-html"><font color="orange"></span> ◦ <span class="raw-html"></font></span></p></td>
<td><p><span class="raw-html"><font color="orange"></span> • <span class="raw-html"></font></span> <span class="raw-html"><font color="orange"></span> ◦ <span class="raw-html"></font></span></p></td>
<td><p><span class="raw-html"><font color="orange"></span> • <span class="raw-html"></font></span> <span class="raw-html"><font color="orange"></span> ◦ <span class="raw-html"></font></span></p></td>
</tr>
<tr class="row-even"><td><p>GoogLeNet</p></td>
<td></td>
<td></td>
<td></td>
<td></td>
<td></td>
<td></td>
<td><p><span class="raw-html"><font color="orange"></span> • <span class="raw-html"></font></span></p></td>
<td><p><span class="raw-html"><font color="orange"></span> • <span class="raw-html"></font></span></p></td>
<td><p><span class="raw-html"><font color="orange"></span> • <span class="raw-html"></font></span></p></td>
</tr>
<tr class="row-odd"><td><p>ResNet-18</p></td>
<td></td>
<td></td>
<td></td>
<td></td>
<td></td>
<td></td>
<td><p><span class="raw-html"><font color="orange"></span> • <span class="raw-html"></font></span></p></td>
<td><p><span class="raw-html"><font color="orange"></span> • <span class="raw-html"></font></span></p></td>
<td><p><span class="raw-html"><font color="orange"></span> • <span class="raw-html"></font></span></p></td>
</tr>
<tr class="row-even"><td><p>ResNet-34</p></td>
<td></td>
<td></td>
<td></td>
<td></td>
<td></td>
<td></td>
<td></td>
<td><p><span class="raw-html"><font color="orange"></span> • <span class="raw-html"></font></span></p></td>
<td><p><span class="raw-html"><font color="orange"></span> • <span class="raw-html"></font></span></p></td>
</tr>
<tr class="row-odd"><td><p>ResNet-50</p></td>
<td></td>
<td></td>
<td></td>
<td></td>
<td></td>
<td></td>
<td></td>
<td></td>
<td><p><span class="raw-html"><font color="orange"></span> ◦ <span class="raw-html"></font></span></p></td>
</tr>
</tbody>
</table>
<dl class="simple">
<dt>Stratix 10</dt><dd><p>Neural networks compatibility table with DNeuro, in terms of memory requirement.</p>
</dd>
</dl>
<table class="docutils align-default">
<colgroup>
<col style="width: 13%" />
<col style="width: 10%" />
<col style="width: 11%" />
<col style="width: 11%" />
<col style="width: 11%" />
<col style="width: 11%" />
<col style="width: 11%" />
<col style="width: 11%" />
<col style="width: 11%" />
</colgroup>
<thead>
<tr class="row-odd"><th class="head" rowspan="2"><p><strong>Stratix 10</strong></p></th>
<th class="head"><p><strong>GX/SX</strong></p></th>
<th class="head"><p><strong>GX/SX</strong></p></th>
<th class="head"><p><strong>GX/SX</strong></p></th>
<th class="head"><p><strong>GX/SX</strong></p></th>
<th class="head"><p><strong>GX/SX</strong></p></th>
<th class="head"><p><strong>GX/SX</strong></p></th>
<th class="head"><p><strong>GX/SX</strong></p></th>
<th class="head"><p><strong>GX/SX</strong></p></th>
</tr>
<tr class="row-even"><th class="head"><p><strong>400</strong></p></th>
<th class="head"><p><strong>650</strong></p></th>
<th class="head"><p><strong>850</strong></p></th>
<th class="head"><p><strong>1100</strong></p></th>
<th class="head"><p><strong>1650</strong></p></th>
<th class="head"><p><strong>2100</strong></p></th>
<th class="head"><p><strong>2500</strong></p></th>
<th class="head"><p><strong>2800</strong></p></th>
</tr>
<tr class="row-odd"><th class="head"><p><strong>M20K (MB)</strong></p></th>
<th class="head"><p>3.75</p></th>
<th class="head"><p>6.12</p></th>
<th class="head"><p>8.5</p></th>
<th class="head"><p>13.37</p></th>
<th class="head"><p>14.25</p></th>
<th class="head"><p>15.87</p></th>
<th class="head"><p>24.37</p></th>
<th class="head"><p>28.62</p></th>
</tr>
<tr class="row-even"><th class="head"><p><strong>DSP</strong></p></th>
<th class="head"><p>648</p></th>
<th class="head"><p>1,152</p></th>
<th class="head"><p>2,016</p></th>
<th class="head"><p>2,592</p></th>
<th class="head"><p>3,145</p></th>
<th class="head"><p>3,744</p></th>
<th class="head"><p>5,011</p></th>
<th class="head"><p>5,760</p></th>
</tr>
<tr class="row-odd"><th class="head"><p><strong>Mult. (MAC/c.)</strong></p></th>
<th class="head"><p>1,296</p></th>
<th class="head"><p>2,304</p></th>
<th class="head"><p>4,032</p></th>
<th class="head"><p>5,184</p></th>
<th class="head"><p>6,290</p></th>
<th class="head"><p>7,488</p></th>
<th class="head"><p>10,022</p></th>
<th class="head"><p>11,520</p></th>
</tr>
</thead>
<tbody>
<tr class="row-even"><td><p>MobileNet_v1_0.25</p></td>
<td><p><span class="math notranslate nohighlight">\(\bullet\)</span></p></td>
<td><p><span class="math notranslate nohighlight">\(\bullet\)</span></p></td>
<td><p><span class="math notranslate nohighlight">\(\bullet\)</span></p></td>
<td><p><span class="math notranslate nohighlight">\(\bullet\)</span></p></td>
<td><p><span class="math notranslate nohighlight">\(\bullet\)</span></p></td>
<td><p><span class="math notranslate nohighlight">\(\bullet\)</span></p></td>
<td><p><span class="math notranslate nohighlight">\(\bullet\)</span></p></td>
<td><p><span class="math notranslate nohighlight">\(\bullet\)</span></p></td>
</tr>
<tr class="row-odd"><td><p>MobileNet_v1_0.5</p></td>
<td><p><span class="math notranslate nohighlight">\(\bullet\)</span></p></td>
<td><p><span class="math notranslate nohighlight">\(\bullet\)</span></p></td>
<td><p><span class="math notranslate nohighlight">\(\bullet\)</span></p></td>
<td><p><span class="math notranslate nohighlight">\(\bullet\)</span></p></td>
<td><p><span class="math notranslate nohighlight">\(\bullet\)</span></p></td>
<td><p><span class="math notranslate nohighlight">\(\bullet\)</span></p></td>
<td><p><span class="math notranslate nohighlight">\(\bullet\)</span></p></td>
<td><p><span class="math notranslate nohighlight">\(\bullet\)</span></p></td>
</tr>
<tr class="row-even"><td><p>MobileNet_v1_0.75</p></td>
<td><p><span class="raw-html"><font color="silver"></span> • <span class="raw-html"></font></span></p></td>
<td><p><span class="math notranslate nohighlight">\(\bullet\)</span></p></td>
<td><p><span class="math notranslate nohighlight">\(\bullet\)</span></p></td>
<td><p><span class="math notranslate nohighlight">\(\bullet\)</span></p></td>
<td><p><span class="math notranslate nohighlight">\(\bullet\)</span></p></td>
<td><p><span class="math notranslate nohighlight">\(\bullet\)</span></p></td>
<td><p><span class="math notranslate nohighlight">\(\bullet\)</span></p></td>
<td><p><span class="math notranslate nohighlight">\(\bullet\)</span></p></td>
</tr>
<tr class="row-odd"><td><p>MobileNet_v1_1.0</p></td>
<td></td>
<td><p><span class="raw-html"><font color="silver"></span> • <span class="raw-html"></font></span></p></td>
<td><p><span class="math notranslate nohighlight">\(\bullet\)</span></p></td>
<td><p><span class="math notranslate nohighlight">\(\bullet\)</span></p></td>
<td><p><span class="math notranslate nohighlight">\(\bullet\)</span></p></td>
<td><p><span class="math notranslate nohighlight">\(\bullet\)</span></p></td>
<td><p><span class="math notranslate nohighlight">\(\bullet\)</span></p></td>
<td><p><span class="math notranslate nohighlight">\(\bullet\)</span></p></td>
</tr>
<tr class="row-even"><td><p>SqueezeNet_v1.0</p></td>
<td><p><span class="raw-html"><font color="orange"></span> • <span class="raw-html"></font></span> <span class="math notranslate nohighlight">\(\bullet\)</span></p></td>
<td><p><span class="math notranslate nohighlight">\(\bullet\)</span></p></td>
<td><p><span class="math notranslate nohighlight">\(\bullet\)</span></p></td>
<td><p><span class="math notranslate nohighlight">\(\bullet\)</span></p></td>
<td><p><span class="math notranslate nohighlight">\(\bullet\)</span></p></td>
<td><p><span class="math notranslate nohighlight">\(\bullet\)</span></p></td>
<td><p><span class="math notranslate nohighlight">\(\bullet\)</span></p></td>
<td><p><span class="math notranslate nohighlight">\(\bullet\)</span></p></td>
</tr>
<tr class="row-odd"><td><p>SqueezeNet_v1.1</p></td>
<td><p><span class="raw-html"><font color="orange"></span> • <span class="raw-html"></font></span> <span class="math notranslate nohighlight">\(\bullet\)</span></p></td>
<td><p><span class="math notranslate nohighlight">\(\bullet\)</span></p></td>
<td><p><span class="math notranslate nohighlight">\(\bullet\)</span></p></td>
<td><p><span class="math notranslate nohighlight">\(\bullet\)</span></p></td>
<td><p><span class="math notranslate nohighlight">\(\bullet\)</span></p></td>
<td><p><span class="math notranslate nohighlight">\(\bullet\)</span></p></td>
<td><p><span class="math notranslate nohighlight">\(\bullet\)</span></p></td>
<td><p><span class="math notranslate nohighlight">\(\bullet\)</span></p></td>
</tr>
<tr class="row-even"><td><p>MobileNet_v2_0.35</p></td>
<td><p><span class="math notranslate nohighlight">\(\bullet\)</span></p></td>
<td><p><span class="math notranslate nohighlight">\(\bullet\)</span></p></td>
<td><p><span class="math notranslate nohighlight">\(\bullet\)</span></p></td>
<td><p><span class="math notranslate nohighlight">\(\bullet\)</span></p></td>
<td><p><span class="math notranslate nohighlight">\(\bullet\)</span></p></td>
<td><p><span class="math notranslate nohighlight">\(\bullet\)</span></p></td>
<td><p><span class="math notranslate nohighlight">\(\bullet\)</span></p></td>
<td><p><span class="math notranslate nohighlight">\(\bullet\)</span></p></td>
</tr>
<tr class="row-odd"><td><p>MobileNet_v2_0.5</p></td>
<td><p><span class="raw-html"><font color="silver"></span> • <span class="raw-html"></font></span></p></td>
<td><p><span class="math notranslate nohighlight">\(\bullet\)</span></p></td>
<td><p><span class="math notranslate nohighlight">\(\bullet\)</span></p></td>
<td><p><span class="math notranslate nohighlight">\(\bullet\)</span></p></td>
<td><p><span class="math notranslate nohighlight">\(\bullet\)</span></p></td>
<td><p><span class="math notranslate nohighlight">\(\bullet\)</span></p></td>
<td><p><span class="math notranslate nohighlight">\(\bullet\)</span></p></td>
<td><p><span class="math notranslate nohighlight">\(\bullet\)</span></p></td>
</tr>
<tr class="row-even"><td><p>MobileNet_v2_0.75</p></td>
<td><p><span class="raw-html"><font color="silver"></span> • <span class="raw-html"></font></span></p></td>
<td><p><span class="math notranslate nohighlight">\(\bullet\)</span></p></td>