1
+
2
+
3
+ <!DOCTYPE html>
4
+ < html class ="writer-html5 " lang ="en " >
5
+ < head >
6
+ < meta charset ="utf-8 " />
7
+
8
+ < meta name ="viewport " content ="width=device-width, initial-scale=1.0 " />
9
+
10
+ < title > YOLOv5: Sparsifying to Improve Object Detection Performance — Neural Magic 0.6.0 documentation</ title >
11
+
12
+
13
+
14
+ < link rel ="stylesheet " href ="../../_static/css/theme.css " type ="text/css " />
15
+ < link rel ="stylesheet " href ="../../_static/pygments.css " type ="text/css " />
16
+ < link rel ="stylesheet " href ="../../_static/pygments.css " type ="text/css " />
17
+ < link rel ="stylesheet " href ="../../_static/css/theme.css " type ="text/css " />
18
+ < link rel ="stylesheet " href ="../../_static/copybutton.css " type ="text/css " />
19
+ < link rel ="stylesheet " href ="../../_static/css/nm-theme-adjustment.css " type ="text/css " />
20
+
21
+
22
+
23
+ < link rel ="shortcut icon " href ="../../_static/favicon.ico "/>
24
+
25
+
26
+
27
+
28
+
29
+
30
+
31
+
32
+ <!--[if lt IE 9]>
33
+ <script src="../../_static/js/html5shiv.min.js"></script>
34
+ <![endif]-->
35
+
36
+
37
+ < script type ="text/javascript " id ="documentation_options " data-url_root ="../../ " src ="../../_static/documentation_options.js "> </ script >
38
+ < script data-url_root ="../../ " id ="documentation_options " src ="../../_static/documentation_options.js "> </ script >
39
+ < script src ="../../_static/jquery.js "> </ script >
40
+ < script src ="../../_static/underscore.js "> </ script >
41
+ < script src ="../../_static/doctools.js "> </ script >
42
+ < script src ="../../_static/clipboard.min.js "> </ script >
43
+ < script src ="../../_static/copybutton.js "> </ script >
44
+
45
+ < script type ="text/javascript " src ="../../_static/js/theme.js "> </ script >
46
+
47
+
48
+ < link rel ="index " title ="Index " href ="../../genindex.html " />
49
+ < link rel ="search " title ="Search " href ="../../search.html " />
50
+ < link rel ="next " title ="YOLOv3: Sparsifying to Improve Object Detection Performance " href ="cv-detection-yolov3.html " />
51
+ < link rel ="prev " title ="MobileNetV1: Sparsifying to Improve Image Classification Performance " href ="cv-classification-mobilenet.html " />
52
+ < script type ="text/javascript " src ="//script.crazyegg.com/pages/scripts/0092/1055.js " async ="async " > </ script >
53
+
54
+ </ head >
55
+
56
+ < body class ="wy-body-for-nav ">
57
+
58
+
59
+ < div class ="wy-grid-for-nav ">
60
+
61
+ < nav data-toggle ="wy-nav-shift " class ="wy-nav-side ">
62
+ < div class ="wy-side-scroll ">
63
+ < div class ="wy-side-nav-search " >
64
+
65
+
66
+
67
+ < a href ="../../index.html " class ="icon icon-home "> Neural Magic
68
+
69
+
70
+
71
+
72
+ < img src ="../../_static/icon-neuralmagic.png " class ="logo " alt ="Logo "/>
73
+
74
+ </ a >
75
+
76
+
77
+
78
+
79
+ < div class ="version ">
80
+ 0.6
81
+ </ div >
82
+
83
+
84
+
85
+
86
+ < div role ="search ">
87
+ < form id ="rtd-search-form " class ="wy-form " action ="../../search.html " method ="get ">
88
+ < input type ="text " name ="q " placeholder ="Search docs " />
89
+ < input type ="hidden " name ="check_keywords " value ="yes " />
90
+ < input type ="hidden " name ="area " value ="default " />
91
+ </ form >
92
+ </ div >
93
+
94
+
95
+ </ div >
96
+
97
+
98
+ < div class ="wy-menu wy-menu-vertical " data-spy ="affix " role ="navigation " aria-label ="main navigation ">
99
+
100
+
101
+
102
+
103
+
104
+
105
+ < p class ="caption "> < span class ="caption-text "> Get Started</ span > </ p >
106
+ < ul >
107
+ < li class ="toctree-l1 "> < a class ="reference internal " href ="../getstarted.html "> Get Started</ a > </ li >
108
+ </ ul >
109
+ < p class ="caption "> < span class ="caption-text "> Sparse Models</ span > </ p >
110
+ < ul class ="current ">
111
+ < li class ="toctree-l1 "> < a class ="reference internal " href ="cv-classification-resnet50.html "> ResNet-50</ a > </ li >
112
+ < li class ="toctree-l1 "> < a class ="reference internal " href ="cv-classification-mobilenet.html "> MobileNetV1</ a > </ li >
113
+ < li class ="toctree-l1 current "> < a class ="current reference internal " href ="# "> YOLOv5</ a > < ul >
114
+ < li class ="toctree-l2 "> < a class ="reference internal " href ="#sparse-models "> Sparse Models</ a > </ li >
115
+ < li class ="toctree-l2 "> < a class ="reference internal " href ="#sparse-inference "> Sparse Inference</ a > </ li >
116
+ < li class ="toctree-l2 "> < a class ="reference internal " href ="#sparse-training "> Sparse Training</ a > </ li >
117
+ </ ul >
118
+ </ li >
119
+ < li class ="toctree-l1 "> < a class ="reference internal " href ="cv-detection-yolov3.html "> YOLOv3</ a > </ li >
120
+ < li class ="toctree-l1 "> < a class ="reference internal " href ="nlp-bert.html "> BERT</ a > </ li >
121
+ </ ul >
122
+ < p class ="caption "> < span class ="caption-text "> Details</ span > </ p >
123
+ < ul >
124
+ < li class ="toctree-l1 "> < a class ="reference external " href ="https://neuralmagic.com/resources/technical-papers "> Technical Papers</ a > </ li >
125
+ < li class ="toctree-l1 "> < a class ="reference internal " href ="../faqs.html "> FAQs</ a > </ li >
126
+ < li class ="toctree-l1 "> < a class ="reference internal " href ="../glossary.html "> Glossary</ a > </ li >
127
+ </ ul >
128
+ < p class ="caption "> < span class ="caption-text "> Connect Online</ span > </ p >
129
+ < ul >
130
+ < li class ="toctree-l1 "> < a class ="reference external " href ="https://github.com/neuralmagic/docs/issues "> Bugs, Feature Requests</ a > </ li >
131
+ < li class ="toctree-l1 "> < a class ="reference external " href ="https://discuss.neuralmagic.com/ "> Support, General Q&A Forums</ a > </ li >
132
+ < li class ="toctree-l1 "> < a class ="reference external " href ="https://join.slack.com/t/discuss-neuralmagic/shared_invite/zt-q1a1cnvo-YBoICSIw3L1dmQpjBeDurQ/ "> Deep Sparse Community</ a > </ li >
133
+ < li class ="toctree-l1 "> < a class ="reference external " href ="https://github.com/neuralmagic "> Neural Magic GitHub</ a > </ li >
134
+ < li class ="toctree-l1 "> < a class ="reference external " href ="https://docs.neuralmagic.com "> Neural Magic Docs</ a > </ li >
135
+ </ ul >
136
+
137
+
138
+
139
+ </ div >
140
+
141
+ </ div >
142
+ </ nav >
143
+
144
+ < section data-toggle ="wy-nav-shift " class ="wy-nav-content-wrap ">
145
+
146
+
147
+ < nav class ="wy-nav-top " aria-label ="top navigation ">
148
+
149
+ < i data-toggle ="wy-nav-top " class ="fa fa-bars "> </ i >
150
+ < a href ="../../index.html "> Neural Magic</ a >
151
+
152
+ </ nav >
153
+
154
+
155
+ < div class ="wy-nav-content ">
156
+
157
+ < div class ="rst-content ">
158
+
159
+
160
+
161
+
162
+
163
+
164
+
165
+
166
+
167
+
168
+
169
+
170
+
171
+
172
+
173
+
174
+
175
+
176
+
177
+ < div role ="navigation " aria-label ="breadcrumbs navigation ">
178
+
179
+ < ul class ="wy-breadcrumbs ">
180
+
181
+ < li > < a href ="../../index.html " class ="icon icon-home "> </ a > »</ li >
182
+
183
+ < li > YOLOv5: Sparsifying to Improve Object Detection Performance</ li >
184
+
185
+
186
+ < li class ="wy-breadcrumbs-aside ">
187
+
188
+
189
+ < a href ="../../_sources/source/model-pages/cv-detection-yolov5.md.txt " rel ="nofollow "> View page source</ a >
190
+
191
+
192
+ </ li >
193
+
194
+ </ ul >
195
+
196
+
197
+ < hr />
198
+ </ div >
199
+ < div role ="main " class ="document " itemscope ="itemscope " itemtype ="http://schema.org/Article ">
200
+ < div itemprop ="articleBody ">
201
+
202
+ <!--
203
+ Copyright (c) 2021 - present / Neuralmagic, Inc. All Rights Reserved.
204
+
205
+ Licensed under the Apache License, Version 2.0 (the "License");
206
+ you may not use this file except in compliance with the License.
207
+ You may obtain a copy of the License at
208
+
209
+ http://www.apache.org/licenses/LICENSE-2.0
210
+
211
+ Unless required by applicable law or agreed to in writing,
212
+ software distributed under the License is distributed on an "AS IS" BASIS,
213
+ WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
214
+ See the License for the specific language governing permissions and
215
+ limitations under the License.
216
+ --> < div class ="section " id ="yolov5-sparsifying-to-improve-object-detection-performance ">
217
+ < h1 > YOLOv5: Sparsifying to Improve Object Detection Performance< a class ="headerlink " href ="#yolov5-sparsifying-to-improve-object-detection-performance " title ="Permalink to this headline "> ¶</ a > </ h1 >
218
+ < p > Neural Magic creates models and recipes that allow anyone to plug in their data and leverage SparseML’s recipe-driven approach on top of Ultralytics’ robust training pipelines for the popular YOLOv5 object detection networks.
219
+ Sparsifying involves removing redundant information from neural networks using algorithms such as pruning and quantization, among others.
220
+ This sparsification process results in faster inference and smaller file sizes for deployments.</ p >
221
+ < p > This page walks through the following use cases for trying out the sparsified YOLOv5 models:</ p >
222
+ < ul class ="simple ">
223
+ < li > < p > Compare the differences between the models for both accuracy and inference performance</ p > </ li >
224
+ < li > < p > Run the models for inference in deployment or applications</ p > </ li >
225
+ < li > < p > Train the models on new datasets</ p > </ li >
226
+ </ ul >
227
+ < div class ="section " id ="sparse-models ">
228
+ < h2 > Sparse Models< a class ="headerlink " href ="#sparse-models " title ="Permalink to this headline "> ¶</ a > </ h2 >
229
+ < div style ="margin-bottom: 24px; display: flex; flex-direction: column; width: auto; ">
230
+ < iframe style ="width: 100%; max-width: 1024px; height: 256px; " src ="https://sparsezoo.neuralmagic.com/widgets/models/model-card/cv%2Fdetection%2Fyolov5-l%2Fpytorch%2Fultralytics%2Fcoco%2Fbase-none " title ="SparseZoo YOLOv5l Available Models " frameborder ="0 " > </ iframe >
231
+ < div style ="margin-top: 8px; ">
232
+ < i > YOLOv5l card not loading? View card < a href ="https://sparsezoo.neuralmagic.com/widgets/models/model-card/cv%2Fdetection%2Fyolov5-l%2Fpytorch%2Fultralytics%2Fcoco%2Fbase-none "> here</ a > .</ i >
233
+ </ div >
234
+ </ div > < div style ="margin-bottom: 24px; display: flex; flex-direction: column; width: auto; ">
235
+ < iframe style ="width: 100%; max-width: 1024px; height: 256px; " src ="https://sparsezoo.neuralmagic.com/widgets/models/model-card/cv%2Fdetection%2Fyolov5-s%2Fpytorch%2Fultralytics%2Fcoco%2Fbase-none " title ="SparseZoo YOLOv5s Available Models " frameborder ="0 " > </ iframe >
236
+ < div style ="margin-top: 8px; ">
237
+ < i > YOLOv5s card not loading? View card < a href ="https://sparsezoo.neuralmagic.com/widgets/models/model-card/cv%2Fdetection%2Fyolov5-s%2Fpytorch%2Fultralytics%2Fcoco%2Fbase-none "> here</ a > .</ i >
238
+ </ div >
239
+ </ div > </ div >
240
+ < div class ="section " id ="sparse-inference ">
241
+ < h2 > Sparse Inference< a class ="headerlink " href ="#sparse-inference " title ="Permalink to this headline "> ¶</ a > </ h2 >
242
+ < div style ="margin-bottom: 24px; display: flex; flex-direction: row; ">
243
+ < a href ="https://github.com/neuralmagic/deepsparse/tree/main/examples/ultralytics-yolo#benchmarking-example " class ="model-page-button " style ="display: flex; flex-direction: column; margin-right: 8px; align-items: center; max-width: 196px ">
244
+ < img src ="https://docs.neuralmagic.com/docs/source/model-pages/images/icon-benchmark.png " alt ="Benchmark " style ="max-width: 160px; width: 100%; " />
245
+ < div style ="text-align: center; "> Benchmark</ div >
246
+ </ a >
247
+ < a href ="https://github.com/neuralmagic/deepsparse/tree/main/examples/ultralytics-yolo#annotation-example " class ="model-page-button " style ="display: flex; flex-direction: column; margin-right: 8px; align-items: center; max-width: 196px ">
248
+ < img src ="https://docs.neuralmagic.com/docs/source/model-pages/images/icon-example-application.png " alt ="Example Application " style ="max-width: 160px; width: 100%; " />
249
+ < div style ="text-align: center; "> Example Application</ div >
250
+ </ a >
251
+ < a href ="https://github.com/neuralmagic/deepsparse/tree/main/examples/ultralytics-yolo#example-yolo-deepsparse-flask-server " class ="model-page-button " style ="display: flex; flex-direction: column; margin-right: 8px; align-items: center; max-width: 196px ">
252
+ < img src ="https://docs.neuralmagic.com/docs/source/model-pages/images/icon-example-deployment.png " alt ="Example Deployment " style ="max-width: 160px; width: 100%; " />
253
+ < div style ="text-align: center; "> Example Deployment</ div >
254
+ </ a >
255
+ < a class ="model-page-button disable-button " style ="display: flex; flex-direction: column; margin-right: 8px; align-items: center; max-width: 196px ">
256
+ < img src ="https://docs.neuralmagic.com/docs/source/model-pages/images/icon-comparison.png " alt ="Performance Comparisons " style ="max-width: 160px; width: 100%; " />
257
+ < div style ="text-align: center; "> Performance Comparisons</ div >
258
+ < div class ="tooltip "> Coming soon!</ div >
259
+ </ a >
260
+ </ div > </ div >
261
+ < div class ="section " id ="sparse-training ">
262
+ < h2 > Sparse Training< a class ="headerlink " href ="#sparse-training " title ="Permalink to this headline "> ¶</ a > </ h2 >
263
+ < div style ="margin-bottom: 24px; display: flex; flex-direction: row; ">
264
+ < a href ="https://github.com/neuralmagic/sparseml/blob/main/integrations/ultralytics-yolov5/tutorials/yolov5_sparse_transfer_learning.md " class ="model-page-button " style ="display: flex; flex-direction: column; margin-right: 8px; align-items: center; max-width: 196px ">
265
+ < img src ="https://docs.neuralmagic.com/docs/source/model-pages/images/icon-train-transfer-learn.png " alt ="Sparse Transfer Learning " style ="max-width: 160px; width: 100%; " />
266
+ < div style ="text-align: center; "> Sparse Transfer Learning</ div >
267
+ </ a >
268
+ < a href ="https://github.com/neuralmagic/sparseml/blob/main/integrations/ultralytics-yolov5/tutorials/sparsifying_yolov5_using_recipes.md " class ="model-page-button " style ="display: flex; flex-direction: column; margin-right: 8px; align-items: center; max-width: 196px ">
269
+ < img src ="https://docs.neuralmagic.com/docs/source/model-pages/images/icon-train-apply-recipe.png " alt ="Applying a Recipe " style ="max-width: 160px; width: 100%; " />
270
+ < div style ="text-align: center; "> Applying a Recipe</ div >
271
+ </ a >
272
+ < a class ="model-page-button disable-button " style ="display: flex; flex-direction: column; margin-right: 8px; align-items: center; max-width: 196px ">
273
+ < img src ="https://docs.neuralmagic.com/docs/source/model-pages/images/icon-train-create-recipe.png " alt ="Creating a Recipe " style ="max-width: 160px; width: 100%; " />
274
+ < div style ="text-align: center; "> Creating a Recipe</ div >
275
+ < div class ="tooltip "> Coming soon!</ div >
276
+ </ a >
277
+ < a style ="pointer-events: none; margin-right: 8px; width: 100%; max-width: 196px; ">
278
+ <!-- placeholder for 4 column grid -->
279
+ </ a >
280
+ </ div > </ div >
281
+ </ div >
282
+
283
+
284
+ </ div >
285
+
286
+ </ div >
287
+ < footer >
288
+ < div class ="rst-footer-buttons " role ="navigation " aria-label ="footer navigation ">
289
+ < a href ="cv-detection-yolov3.html " class ="btn btn-neutral float-right " title ="YOLOv3: Sparsifying to Improve Object Detection Performance " accesskey ="n " rel ="next "> Next < span class ="fa fa-arrow-circle-right " aria-hidden ="true "> </ span > </ a >
290
+ < a href ="cv-classification-mobilenet.html " class ="btn btn-neutral float-left " title ="MobileNetV1: Sparsifying to Improve Image Classification Performance " accesskey ="p " rel ="prev "> < span class ="fa fa-arrow-circle-left " aria-hidden ="true "> </ span > Previous</ a >
291
+ </ div >
292
+
293
+ < hr />
294
+
295
+ < div role ="contentinfo ">
296
+ < p >
297
+ © Copyright 2021 - present / Neuralmagic, Inc. All Rights Reserved / < a href ="https://neuralmagic.com/legal "> Legal</ a >
298
+
299
+ </ p >
300
+ </ div >
301
+ </ footer >
302
+ </ div >
303
+ </ div >
304
+
305
+ </ section >
306
+
307
+ </ div >
308
+
309
+ < div class ="rst-versions " data-toggle ="rst-versions " role ="note " aria-label ="versions ">
310
+ < span class ="rst-current-version " data-toggle ="rst-current-version ">
311
+ < span class ="fa fa-book "> Other Versions</ span >
312
+ v: main
313
+ < span class ="fa fa-caret-down "> </ span >
314
+ </ span >
315
+ < div class ="rst-other-versions ">
316
+ < dl >
317
+ < dt > Tags</ dt >
318
+ < dd > < a href ="../../../v.0.3.0/index.html "> v.0.3.0</ a > </ dd >
319
+ < dd > < a href ="../../../v0.3.1/index.html "> v0.3.1</ a > </ dd >
320
+ < dd > < a href ="../../../v0.4.0/index.html "> v0.4.0</ a > </ dd >
321
+ < dd > < a href ="../../../v0.5.0/index.html "> v0.5.0</ a > </ dd >
322
+ < dd > < a href ="../../../v0.5.1/index.html "> v0.5.1</ a > </ dd >
323
+ </ dl >
324
+ < dl >
325
+ < dt > Branches</ dt >
326
+ < dd > < a href ="cv-detection-yolov5.html "> main</ a > </ dd >
327
+ </ dl >
328
+ </ div >
329
+ </ div >
330
+
331
+ < script type ="text/javascript ">
332
+ jQuery ( function ( ) {
333
+ SphinxRtdTheme . Navigation . enable ( true ) ;
334
+ } ) ;
335
+ </ script >
336
+
337
+
338
+
339
+
340
+ <!-- Theme Analytics -->
341
+ < script >
342
+ ( function ( i , s , o , g , r , a , m ) { i [ 'GoogleAnalyticsObject' ] = r ; i [ r ] = i [ r ] || function ( ) {
343
+ ( i [ r ] . q = i [ r ] . q || [ ] ) . push ( arguments ) } , i [ r ] . l = 1 * new Date ( ) ; a = s . createElement ( o ) ,
344
+ m = s . getElementsByTagName ( o ) [ 0 ] ; a . async = 1 ; a . src = g ; m . parentNode . insertBefore ( a , m )
345
+ } ) ( window , document , 'script' , 'https://www.google-analytics.com/analytics.js' , 'ga' ) ;
346
+
347
+ ga ( 'create' , 'UA-128364174-1' , 'auto' ) ;
348
+
349
+ ga ( 'send' , 'pageview' ) ;
350
+ </ script >
351
+
352
+
353
+
354
+
355
+ </ body >
356
+ </ html >
0 commit comments