Deci-AI
/
super-gradients
connected to https://github.com/Deci-AI/super-gradients.git


  
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

	
            <!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>super_gradients.training.losses.yolo_v5_loss &mdash; SuperGradients 1.0 documentation</title>
      <link rel="stylesheet" href="../../../../_static/pygments.css" type="text/css" />
      <link rel="stylesheet" href="../../../../_static/css/theme.css" type="text/css" />
      <link rel="stylesheet" href="../../../../_static/graphviz.css" type="text/css" />
  <!--[if lt IE 9]>
    <script src="../../../../_static/js/html5shiv.min.js"></script>
  <![endif]-->
  
        <script data-url_root="../../../../" id="documentation_options" 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/js/theme.js"></script>
    <link rel="index" title="Index" href="../../../../genindex.html" />
    <link rel="search" title="Search" href="../../../../search.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"> SuperGradients
          </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="Navigation menu">
              <p class="caption"><span class="caption-text">Welcome To SuperGradients</span></p>
<ul>
<li class="toctree-l1"><a class="reference internal" href="../../../../welcome.html">SuperGradients</a></li>
</ul>
<p class="caption"><span class="caption-text">Technical Documentation</span></p>
<ul>
<li class="toctree-l1"><a class="reference internal" href="../../../../super_gradients.common.html">Common package</a></li>
<li class="toctree-l1"><a class="reference internal" href="../../../../super_gradients.training.html">Training package</a></li>
</ul>
<p class="caption"><span class="caption-text">User Guide</span></p>
<ul>
<li class="toctree-l1"><a class="reference internal" href="../../../../user_guide.html">What is SuperGradients?</a></li>
<li class="toctree-l1"><a class="reference internal" href="../../../../user_guide.html#introducing-the-supergradients-library">Introducing the SuperGradients library</a></li>
<li class="toctree-l1"><a class="reference internal" href="../../../../user_guide.html#installation">Installation</a></li>
<li class="toctree-l1"><a class="reference internal" href="../../../../user_guide.html#integrating-your-training-code-complete-walkthrough">Integrating Your Training Code - Complete Walkthrough</a></li>
<li class="toctree-l1"><a class="reference internal" href="../../../../user_guide.html#training-parameters">Training Parameters</a></li>
<li class="toctree-l1"><a class="reference internal" href="../../../../user_guide.html#logs-and-checkpoints">Logs and Checkpoints</a></li>
<li class="toctree-l1"><a class="reference internal" href="../../../../user_guide.html#dataset-parameters">Dataset Parameters</a></li>
<li class="toctree-l1"><a class="reference internal" href="../../../../user_guide.html#network-architectures">Network Architectures</a></li>
<li class="toctree-l1"><a class="reference internal" href="../../../../user_guide.html#pretrained-models">Pretrained Models</a></li>
<li class="toctree-l1"><a class="reference internal" href="../../../../user_guide.html#how-to-reproduce-our-training-recipes">How To Reproduce Our Training Recipes</a></li>
<li class="toctree-l1"><a class="reference internal" href="../../../../user_guide.html#supergradients-faq">SuperGradients FAQ</a></li>
</ul>

        </div>
      </div>
    </nav>

    <section data-toggle="wy-nav-shift" class="wy-nav-content-wrap"><nav class="wy-nav-top" aria-label="Mobile navigation menu" >
          <i data-toggle="wy-nav-top" class="fa fa-bars"></i>
          <a href="../../../../index.html">SuperGradients</a>
      </nav>

      <div class="wy-nav-content">
        <div class="rst-content">
          <div role="navigation" aria-label="Page navigation">
  <ul class="wy-breadcrumbs">
      <li><a href="../../../../index.html" class="icon icon-home"></a> &raquo;</li>
          <li><a href="../../../index.html">Module code</a> &raquo;</li>
      <li>super_gradients.training.losses.yolo_v5_loss</li>
      <li class="wy-breadcrumbs-aside">
      </li>
  </ul>
  <hr/>
</div>
          <div role="main" class="document" itemscope="itemscope" itemtype="http://schema.org/Article">
           <div itemprop="articleBody">
             
  <h1>Source code for super_gradients.training.losses.yolo_v5_loss</h1><div class="highlight"><pre>
<span></span><span class="kn">from</span> <span class="nn">typing</span> <span class="kn">import</span> <span class="n">List</span><span class="p">,</span> <span class="n">Tuple</span><span class="p">,</span> <span class="n">Union</span>

<span class="kn">import</span> <span class="nn">torch</span>
<span class="kn">from</span> <span class="nn">torch</span> <span class="kn">import</span> <span class="n">nn</span>
<span class="kn">from</span> <span class="nn">torch.nn.modules.loss</span> <span class="kn">import</span> <span class="n">_Loss</span>

<span class="kn">from</span> <span class="nn">super_gradients.training.losses.focal_loss</span> <span class="kn">import</span> <span class="n">FocalLoss</span>
<span class="kn">from</span> <span class="nn">super_gradients.training.utils.detection_utils</span> <span class="kn">import</span> <span class="n">calculate_bbox_iou_elementwise</span><span class="p">,</span> <span class="n">Anchors</span>


<div class="viewcode-block" id="YoLoV5DetectionLoss"><a class="viewcode-back" href="../../../../super_gradients.training.losses.html#super_gradients.training.losses.YoLoV5DetectionLoss">[docs]</a><span class="k">class</span> <span class="nc">YoLoV5DetectionLoss</span><span class="p">(</span><span class="n">_Loss</span><span class="p">):</span>
    <span class="sd">&quot;&quot;&quot;</span>
<span class="sd">    Calculate YOLO V5 loss:</span>
<span class="sd">    L = L_objectivness + L_boxes + L_classification</span>
<span class="sd">    &quot;&quot;&quot;</span>

    <span class="k">def</span> <span class="fm">__init__</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">anchors</span><span class="p">:</span> <span class="n">Anchors</span><span class="p">,</span>
                 <span class="n">cls_pos_weight</span><span class="p">:</span> <span class="n">Union</span><span class="p">[</span><span class="nb">float</span><span class="p">,</span> <span class="n">List</span><span class="p">[</span><span class="nb">float</span><span class="p">]]</span> <span class="o">=</span> <span class="mf">1.0</span><span class="p">,</span> <span class="n">obj_pos_weight</span><span class="p">:</span> <span class="nb">float</span> <span class="o">=</span> <span class="mf">1.0</span><span class="p">,</span>
                 <span class="n">obj_loss_gain</span><span class="p">:</span> <span class="nb">float</span> <span class="o">=</span> <span class="mf">1.0</span><span class="p">,</span> <span class="n">box_loss_gain</span><span class="p">:</span> <span class="nb">float</span> <span class="o">=</span> <span class="mf">0.05</span><span class="p">,</span> <span class="n">cls_loss_gain</span><span class="p">:</span> <span class="nb">float</span> <span class="o">=</span> <span class="mf">0.5</span><span class="p">,</span>
                 <span class="n">focal_loss_gamma</span><span class="p">:</span> <span class="nb">float</span> <span class="o">=</span> <span class="mf">0.0</span><span class="p">,</span>
                 <span class="n">cls_objectness_weights</span><span class="p">:</span> <span class="n">Union</span><span class="p">[</span><span class="n">List</span><span class="p">[</span><span class="nb">float</span><span class="p">],</span> <span class="n">torch</span><span class="o">.</span><span class="n">Tensor</span><span class="p">]</span> <span class="o">=</span> <span class="kc">None</span><span class="p">,</span> <span class="n">anchor_threshold</span><span class="o">=</span><span class="mf">4.0</span><span class="p">):</span>
        <span class="sd">&quot;&quot;&quot;</span>
<span class="sd">        :param anchors:                 the anchors of the model (same anchors used for training)</span>
<span class="sd">        :param cls_pos_weight:          pos_weight for BCE in L_classification,</span>
<span class="sd">                                        can be one value for all positives or a list of weights for each class</span>
<span class="sd">        :param obj_pos_weight:          pos_weight for BCE in L_objectivness</span>
<span class="sd">        :param obj_loss_gain:           coef for L_objectivness</span>
<span class="sd">        :param box_loss_gain:           coef for L_boxes</span>
<span class="sd">        :param cls_loss_gain:           coef for L_classification</span>
<span class="sd">        :param focal_loss_gamma:        gamma for a focal loss, 0 to train with a usual BCE</span>
<span class="sd">        :param cls_objectness_weights:  class-based weight for L_objectivness that will be applied in each cell that</span>
<span class="sd">                                        has a GT assigned to it.</span>
<span class="sd">                                        Note: default weight for objectness loss in each cell is 1.</span>
<span class="sd">        :param anchor_threshold:                ratio defining a size range of an appropriate anchor.</span>
<span class="sd">        &quot;&quot;&quot;</span>
        <span class="nb">super</span><span class="p">(</span><span class="n">YoLoV5DetectionLoss</span><span class="p">,</span> <span class="bp">self</span><span class="p">)</span><span class="o">.</span><span class="fm">__init__</span><span class="p">()</span>

        <span class="bp">self</span><span class="o">.</span><span class="n">cls_pos_weight</span> <span class="o">=</span> <span class="n">cls_pos_weight</span>
        <span class="bp">self</span><span class="o">.</span><span class="n">obj_pos_weight</span> <span class="o">=</span> <span class="n">obj_pos_weight</span>
        <span class="bp">self</span><span class="o">.</span><span class="n">obj_loss_gain</span> <span class="o">=</span> <span class="n">obj_loss_gain</span>
        <span class="bp">self</span><span class="o">.</span><span class="n">box_loss_gain</span> <span class="o">=</span> <span class="n">box_loss_gain</span>
        <span class="bp">self</span><span class="o">.</span><span class="n">cls_loss_gain</span> <span class="o">=</span> <span class="n">cls_loss_gain</span>
        <span class="bp">self</span><span class="o">.</span><span class="n">focal_loss_gamma</span> <span class="o">=</span> <span class="n">focal_loss_gamma</span>

        <span class="bp">self</span><span class="o">.</span><span class="n">anchor_threshold</span> <span class="o">=</span> <span class="n">anchor_threshold</span>

        <span class="bp">self</span><span class="o">.</span><span class="n">anchors</span> <span class="o">=</span> <span class="n">anchors</span>

        <span class="bp">self</span><span class="o">.</span><span class="n">cls_obj_weights</span> <span class="o">=</span> <span class="n">cls_objectness_weights</span>
        <span class="k">if</span> <span class="nb">isinstance</span><span class="p">(</span><span class="n">cls_objectness_weights</span><span class="p">,</span> <span class="nb">list</span><span class="p">):</span>
            <span class="bp">self</span><span class="o">.</span><span class="n">cls_obj_weights</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">nn</span><span class="o">.</span><span class="n">Parameter</span><span class="p">(</span><span class="n">torch</span><span class="o">.</span><span class="n">tensor</span><span class="p">(</span><span class="n">cls_objectness_weights</span><span class="p">))</span>

<div class="viewcode-block" id="YoLoV5DetectionLoss.forward"><a class="viewcode-back" href="../../../../super_gradients.training.losses.html#super_gradients.training.losses.YoLoV5DetectionLoss.forward">[docs]</a>    <span class="k">def</span> <span class="nf">forward</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">model_output</span><span class="p">,</span> <span class="n">targets</span><span class="p">):</span>
        <span class="k">if</span> <span class="nb">isinstance</span><span class="p">(</span><span class="n">model_output</span><span class="p">,</span> <span class="nb">tuple</span><span class="p">)</span> <span class="ow">and</span> <span class="nb">len</span><span class="p">(</span><span class="n">model_output</span><span class="p">)</span> <span class="o">==</span> <span class="mi">2</span><span class="p">:</span>
            <span class="c1"># in test/eval mode the Yolo v5 model output a tuple where the second item is the raw predictions</span>
            <span class="n">_</span><span class="p">,</span> <span class="n">predictions</span> <span class="o">=</span> <span class="n">model_output</span>
        <span class="k">else</span><span class="p">:</span>
            <span class="n">predictions</span> <span class="o">=</span> <span class="n">model_output</span>

        <span class="k">return</span> <span class="bp">self</span><span class="o">.</span><span class="n">compute_loss</span><span class="p">(</span><span class="n">predictions</span><span class="p">,</span> <span class="n">targets</span><span class="p">)</span></div>

<div class="viewcode-block" id="YoLoV5DetectionLoss.build_targets"><a class="viewcode-back" href="../../../../super_gradients.training.losses.html#super_gradients.training.losses.YoLoV5DetectionLoss.build_targets">[docs]</a>    <span class="k">def</span> <span class="nf">build_targets</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">predictions</span><span class="p">:</span> <span class="n">List</span><span class="p">[</span><span class="n">torch</span><span class="o">.</span><span class="n">Tensor</span><span class="p">],</span> <span class="n">targets</span><span class="p">:</span> <span class="n">torch</span><span class="o">.</span><span class="n">Tensor</span><span class="p">)</span> \
            <span class="o">-&gt;</span> <span class="n">Tuple</span><span class="p">[</span><span class="n">List</span><span class="p">[</span><span class="n">torch</span><span class="o">.</span><span class="n">Tensor</span><span class="p">],</span> <span class="n">List</span><span class="p">[</span><span class="n">torch</span><span class="o">.</span><span class="n">Tensor</span><span class="p">],</span> <span class="n">List</span><span class="p">[</span><span class="n">Tuple</span><span class="p">[</span><span class="n">torch</span><span class="o">.</span><span class="n">Tensor</span><span class="p">]],</span> <span class="n">List</span><span class="p">[</span><span class="n">torch</span><span class="o">.</span><span class="n">Tensor</span><span class="p">]]:</span>
        <span class="sd">&quot;&quot;&quot;</span>
<span class="sd">        Assign targets to anchors to use in L_boxes &amp; L_classification calculation:</span>
<span class="sd">            * each target can be assigned to a few anchors,</span>
<span class="sd">            all anchors that are within [1/self.anchor_threshold, self.anchor_threshold] times target size range</span>
<span class="sd">            * each anchor can be assigned to a few targets</span>

<span class="sd">        :param predictions:         Yolo predictions</span>
<span class="sd">        :param targets:             ground truth targets</span>
<span class="sd">        :return:                    each of 4 outputs contains one element for each Yolo output,</span>
<span class="sd">                                    correspondences are raveled over the whole batch and all anchors:</span>
<span class="sd">                                        * classes of the targets;</span>
<span class="sd">                                        * boxes of the targets;</span>
<span class="sd">                                        * image id in a batch, anchor id, grid y, grid x coordinates;</span>
<span class="sd">                                        * anchor sizes.</span>
<span class="sd">                                    All the above can be indexed in parallel to get the selected correspondences</span>
<span class="sd">        &quot;&quot;&quot;</span>
        <span class="n">num_anchors</span><span class="p">,</span> <span class="n">num_targets</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">anchors</span><span class="o">.</span><span class="n">num_anchors</span><span class="p">,</span> <span class="n">targets</span><span class="o">.</span><span class="n">shape</span><span class="p">[</span><span class="mi">0</span><span class="p">]</span>
        <span class="n">target_classes</span><span class="p">,</span> <span class="n">target_boxes</span><span class="p">,</span> <span class="n">indices</span><span class="p">,</span> <span class="n">anchors</span> <span class="o">=</span> <span class="p">[],</span> <span class="p">[],</span> <span class="p">[],</span> <span class="p">[]</span>

        <span class="n">gain</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">ones</span><span class="p">(</span><span class="mi">7</span><span class="p">,</span> <span class="n">device</span><span class="o">=</span><span class="n">targets</span><span class="o">.</span><span class="n">device</span><span class="p">)</span>  <span class="c1"># normalized to gridspace gain</span>
        <span class="n">anchor_indices</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">arange</span><span class="p">(</span><span class="n">num_anchors</span><span class="p">,</span> <span class="n">device</span><span class="o">=</span><span class="n">targets</span><span class="o">.</span><span class="n">device</span><span class="p">)</span>
        <span class="n">anchor_indices</span> <span class="o">=</span> <span class="n">anchor_indices</span><span class="o">.</span><span class="n">float</span><span class="p">()</span><span class="o">.</span><span class="n">view</span><span class="p">(</span><span class="n">num_anchors</span><span class="p">,</span> <span class="mi">1</span><span class="p">)</span><span class="o">.</span><span class="n">repeat</span><span class="p">(</span><span class="mi">1</span><span class="p">,</span> <span class="n">num_targets</span><span class="p">)</span>
        <span class="c1"># repeat all targets for each anchor and append a corresponding anchor index</span>
        <span class="n">targets</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">cat</span><span class="p">((</span><span class="n">targets</span><span class="o">.</span><span class="n">repeat</span><span class="p">(</span><span class="n">num_anchors</span><span class="p">,</span> <span class="mi">1</span><span class="p">,</span> <span class="mi">1</span><span class="p">),</span> <span class="n">anchor_indices</span><span class="p">[:,</span> <span class="p">:,</span> <span class="kc">None</span><span class="p">]),</span> <span class="mi">2</span><span class="p">)</span>

        <span class="n">bias</span> <span class="o">=</span> <span class="mf">0.5</span>
        <span class="n">off</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">tensor</span><span class="p">([[</span><span class="mi">0</span><span class="p">,</span> <span class="mi">0</span><span class="p">],</span>
                            <span class="p">[</span><span class="mi">1</span><span class="p">,</span> <span class="mi">0</span><span class="p">],</span> <span class="p">[</span><span class="mi">0</span><span class="p">,</span> <span class="mi">1</span><span class="p">],</span> <span class="p">[</span><span class="o">-</span><span class="mi">1</span><span class="p">,</span> <span class="mi">0</span><span class="p">],</span> <span class="p">[</span><span class="mi">0</span><span class="p">,</span> <span class="o">-</span><span class="mi">1</span><span class="p">],</span>  <span class="c1"># j,k,l,m</span>
                            <span class="p">],</span> <span class="n">device</span><span class="o">=</span><span class="n">targets</span><span class="o">.</span><span class="n">device</span><span class="p">)</span><span class="o">.</span><span class="n">float</span><span class="p">()</span> <span class="o">*</span> <span class="n">bias</span>  <span class="c1"># offsets</span>

        <span class="k">for</span> <span class="n">i</span> <span class="ow">in</span> <span class="nb">range</span><span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">anchors</span><span class="o">.</span><span class="n">detection_layers_num</span><span class="p">):</span>
            <span class="n">anch</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">anchors</span><span class="o">.</span><span class="n">anchors</span><span class="p">[</span><span class="n">i</span><span class="p">]</span>
            <span class="n">gain</span><span class="p">[</span><span class="mi">2</span><span class="p">:</span><span class="mi">6</span><span class="p">]</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">tensor</span><span class="p">(</span><span class="n">predictions</span><span class="p">[</span><span class="n">i</span><span class="p">]</span><span class="o">.</span><span class="n">shape</span><span class="p">)[[</span><span class="mi">3</span><span class="p">,</span> <span class="mi">2</span><span class="p">,</span> <span class="mi">3</span><span class="p">,</span> <span class="mi">2</span><span class="p">]]</span>  <span class="c1"># xyxy gain</span>

            <span class="c1"># Convert target coordinates from [0, 1] range to coordinates in [0, GridY], [0, GridX] ranges</span>
            <span class="n">t</span> <span class="o">=</span> <span class="n">targets</span> <span class="o">*</span> <span class="n">gain</span>
            <span class="k">if</span> <span class="n">num_targets</span><span class="p">:</span>
                <span class="c1"># Match: filter targets by anchor size ratio</span>
                <span class="n">r</span> <span class="o">=</span> <span class="n">t</span><span class="p">[:,</span> <span class="p">:,</span> <span class="mi">4</span><span class="p">:</span><span class="mi">6</span><span class="p">]</span> <span class="o">/</span> <span class="n">anch</span><span class="p">[:,</span> <span class="kc">None</span><span class="p">]</span>  <span class="c1"># wh ratio</span>
                <span class="n">filtered_targets_ids</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">max</span><span class="p">(</span><span class="n">r</span><span class="p">,</span> <span class="mf">1.</span> <span class="o">/</span> <span class="n">r</span><span class="p">)</span><span class="o">.</span><span class="n">max</span><span class="p">(</span><span class="mi">2</span><span class="p">)[</span><span class="mi">0</span><span class="p">]</span> <span class="o">&lt;</span> <span class="bp">self</span><span class="o">.</span><span class="n">anchor_threshold</span>  <span class="c1"># compare</span>
                <span class="n">t</span> <span class="o">=</span> <span class="n">t</span><span class="p">[</span><span class="n">filtered_targets_ids</span><span class="p">]</span>

                <span class="c1"># Find coordinates of targets on a grid</span>
                <span class="n">gxy</span> <span class="o">=</span> <span class="n">t</span><span class="p">[:,</span> <span class="mi">2</span><span class="p">:</span><span class="mi">4</span><span class="p">]</span>  <span class="c1"># grid xy</span>
                <span class="n">gxi</span> <span class="o">=</span> <span class="n">gain</span><span class="p">[[</span><span class="mi">2</span><span class="p">,</span> <span class="mi">3</span><span class="p">]]</span> <span class="o">-</span> <span class="n">gxy</span>  <span class="c1"># inverse</span>
                <span class="n">j</span><span class="p">,</span> <span class="n">k</span> <span class="o">=</span> <span class="p">((</span><span class="n">gxy</span> <span class="o">%</span> <span class="mf">1.</span> <span class="o">&lt;</span> <span class="n">bias</span><span class="p">)</span> <span class="o">&amp;</span> <span class="p">(</span><span class="n">gxy</span> <span class="o">&gt;</span> <span class="mf">1.</span><span class="p">))</span><span class="o">.</span><span class="n">T</span>
                <span class="n">l</span><span class="p">,</span> <span class="n">m</span> <span class="o">=</span> <span class="p">((</span><span class="n">gxi</span> <span class="o">%</span> <span class="mf">1.</span> <span class="o">&lt;</span> <span class="n">bias</span><span class="p">)</span> <span class="o">&amp;</span> <span class="p">(</span><span class="n">gxi</span> <span class="o">&gt;</span> <span class="mf">1.</span><span class="p">))</span><span class="o">.</span><span class="n">T</span>
                <span class="n">j</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">stack</span><span class="p">((</span><span class="n">torch</span><span class="o">.</span><span class="n">ones_like</span><span class="p">(</span><span class="n">j</span><span class="p">),</span> <span class="n">j</span><span class="p">,</span> <span class="n">k</span><span class="p">,</span> <span class="n">l</span><span class="p">,</span> <span class="n">m</span><span class="p">))</span>
                <span class="n">t</span> <span class="o">=</span> <span class="n">t</span><span class="o">.</span><span class="n">repeat</span><span class="p">((</span><span class="mi">5</span><span class="p">,</span> <span class="mi">1</span><span class="p">,</span> <span class="mi">1</span><span class="p">))[</span><span class="n">j</span><span class="p">]</span>
                <span class="n">offsets</span> <span class="o">=</span> <span class="p">(</span><span class="n">torch</span><span class="o">.</span><span class="n">zeros_like</span><span class="p">(</span><span class="n">gxy</span><span class="p">)[</span><span class="kc">None</span><span class="p">]</span> <span class="o">+</span> <span class="n">off</span><span class="p">[:,</span> <span class="kc">None</span><span class="p">])[</span><span class="n">j</span><span class="p">]</span>
            <span class="k">else</span><span class="p">:</span>
                <span class="n">t</span> <span class="o">=</span> <span class="n">targets</span><span class="p">[</span><span class="mi">0</span><span class="p">]</span>
                <span class="n">offsets</span> <span class="o">=</span> <span class="mi">0</span>

            <span class="c1"># Define</span>
            <span class="n">b</span><span class="p">,</span> <span class="n">c</span> <span class="o">=</span> <span class="n">t</span><span class="p">[:,</span> <span class="p">:</span><span class="mi">2</span><span class="p">]</span><span class="o">.</span><span class="n">long</span><span class="p">()</span><span class="o">.</span><span class="n">T</span>  <span class="c1"># image, class</span>
            <span class="n">gxy</span> <span class="o">=</span> <span class="n">t</span><span class="p">[:,</span> <span class="mi">2</span><span class="p">:</span><span class="mi">4</span><span class="p">]</span>  <span class="c1"># grid xy</span>
            <span class="n">gwh</span> <span class="o">=</span> <span class="n">t</span><span class="p">[:,</span> <span class="mi">4</span><span class="p">:</span><span class="mi">6</span><span class="p">]</span>  <span class="c1"># grid wh</span>
            <span class="n">gij</span> <span class="o">=</span> <span class="p">(</span><span class="n">gxy</span> <span class="o">-</span> <span class="n">offsets</span><span class="p">)</span><span class="o">.</span><span class="n">long</span><span class="p">()</span>
            <span class="n">gi</span><span class="p">,</span> <span class="n">gj</span> <span class="o">=</span> <span class="n">gij</span><span class="o">.</span><span class="n">T</span>  <span class="c1"># grid xy indices</span>
            <span class="c1"># prevent coordinates from going out of bounds</span>
            <span class="n">gi</span><span class="p">,</span> <span class="n">gj</span> <span class="o">=</span> <span class="n">gi</span><span class="o">.</span><span class="n">clamp_</span><span class="p">(</span><span class="mi">0</span><span class="p">,</span> <span class="n">gain</span><span class="p">[</span><span class="mi">2</span><span class="p">]</span> <span class="o">-</span> <span class="mi">1</span><span class="p">),</span> <span class="n">gj</span><span class="o">.</span><span class="n">clamp_</span><span class="p">(</span><span class="mi">0</span><span class="p">,</span> <span class="n">gain</span><span class="p">[</span><span class="mi">3</span><span class="p">]</span> <span class="o">-</span> <span class="mi">1</span><span class="p">)</span>

            <span class="c1"># Append</span>
            <span class="n">a</span> <span class="o">=</span> <span class="n">t</span><span class="p">[:,</span> <span class="mi">6</span><span class="p">]</span><span class="o">.</span><span class="n">long</span><span class="p">()</span>  <span class="c1"># anchor indices</span>
            <span class="n">indices</span><span class="o">.</span><span class="n">append</span><span class="p">((</span><span class="n">b</span><span class="p">,</span> <span class="n">a</span><span class="p">,</span> <span class="n">gj</span><span class="p">,</span> <span class="n">gi</span><span class="p">))</span>  <span class="c1"># image, anchor, grid indices</span>
            <span class="n">target_boxes</span><span class="o">.</span><span class="n">append</span><span class="p">(</span><span class="n">torch</span><span class="o">.</span><span class="n">cat</span><span class="p">((</span><span class="n">gxy</span> <span class="o">-</span> <span class="n">gij</span><span class="p">,</span> <span class="n">gwh</span><span class="p">),</span> <span class="mi">1</span><span class="p">))</span>  <span class="c1"># box</span>
            <span class="n">anchors</span><span class="o">.</span><span class="n">append</span><span class="p">(</span><span class="n">anch</span><span class="p">[</span><span class="n">a</span><span class="p">])</span>  <span class="c1"># anchors</span>
            <span class="n">target_classes</span><span class="o">.</span><span class="n">append</span><span class="p">(</span><span class="n">c</span><span class="p">)</span>  <span class="c1"># class</span>

        <span class="k">return</span> <span class="n">target_classes</span><span class="p">,</span> <span class="n">target_boxes</span><span class="p">,</span> <span class="n">indices</span><span class="p">,</span> <span class="n">anchors</span></div>

<div class="viewcode-block" id="YoLoV5DetectionLoss.compute_loss"><a class="viewcode-back" href="../../../../super_gradients.training.losses.html#super_gradients.training.losses.YoLoV5DetectionLoss.compute_loss">[docs]</a>    <span class="k">def</span> <span class="nf">compute_loss</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">predictions</span><span class="p">:</span> <span class="n">List</span><span class="p">[</span><span class="n">torch</span><span class="o">.</span><span class="n">Tensor</span><span class="p">],</span> <span class="n">targets</span><span class="p">:</span> <span class="n">torch</span><span class="o">.</span><span class="n">Tensor</span><span class="p">,</span> <span class="n">giou_loss_ratio</span><span class="p">:</span> <span class="nb">float</span> <span class="o">=</span> <span class="mf">1.0</span><span class="p">)</span> \
            <span class="o">-&gt;</span> <span class="n">Tuple</span><span class="p">[</span><span class="n">torch</span><span class="o">.</span><span class="n">Tensor</span><span class="p">,</span> <span class="n">torch</span><span class="o">.</span><span class="n">Tensor</span><span class="p">]:</span>
        <span class="sd">&quot;&quot;&quot;</span>
<span class="sd">        L = L_objectivness + L_boxes + L_classification</span>
<span class="sd">        where:</span>
<span class="sd">            * L_boxes and L_classification are calculated only between anchors and targets that suit them;</span>
<span class="sd">            * L_objectivness is calculated on all anchors.</span>

<span class="sd">        L_classification:</span>
<span class="sd">            for anchors that have suitable ground truths in their grid locations add BCEs</span>
<span class="sd">            to force max probability for each GT class in a multi-label way</span>
<span class="sd">            Coef: self.cls_loss_gain</span>
<span class="sd">        L_boxes:</span>
<span class="sd">            for anchors that have suitable ground truths in their grid locations</span>
<span class="sd">            add (1 - IoU), IoU between a predicted box and each GT box, force maximum IoU</span>
<span class="sd">            Coef: self.box_loss_gain</span>
<span class="sd">        L_objectness:</span>
<span class="sd">            for each anchor add BCE to force a prediction of (1 - giou_loss_ratio) + giou_loss_ratio * IoU,</span>
<span class="sd">            IoU between a predicted box and random GT in it</span>
<span class="sd">            Coef: self.obj_loss_gain, loss from each YOLO grid is additionally multiplied by balance = [4.0, 1.0, 0.4]</span>
<span class="sd">                  to balance different contributions coming from different numbers of grid cells</span>

<span class="sd">        :param predictions:     output from all Yolo levels, each of shape</span>
<span class="sd">                                [Batch x Num_Anchors x GridSizeY x GridSizeX x (4 + 1 + Num_classes)]</span>
<span class="sd">        :param targets:         [Num_targets x (4 + 2)], values on dim 1 are: image id in a batch, class, box x y w h</span>
<span class="sd">        :param giou_loss_ratio: a coef in L_objectness defining what should be predicted as objecness</span>
<span class="sd">                                in a call with a target: can be a value in [IoU, 1] range</span>
<span class="sd">        :return:                loss, all losses separately in a detached tensor</span>
<span class="sd">        &quot;&quot;&quot;</span>
        <span class="n">device</span> <span class="o">=</span> <span class="n">targets</span><span class="o">.</span><span class="n">device</span>
        <span class="n">loss_classification</span><span class="p">,</span> <span class="n">loss_boxes</span><span class="p">,</span> <span class="n">loss_objectivness</span> <span class="o">=</span> \
            <span class="n">torch</span><span class="o">.</span><span class="n">zeros</span><span class="p">(</span><span class="mi">1</span><span class="p">,</span> <span class="n">device</span><span class="o">=</span><span class="n">device</span><span class="p">),</span> <span class="n">torch</span><span class="o">.</span><span class="n">zeros</span><span class="p">(</span><span class="mi">1</span><span class="p">,</span> <span class="n">device</span><span class="o">=</span><span class="n">device</span><span class="p">),</span> <span class="n">torch</span><span class="o">.</span><span class="n">zeros</span><span class="p">(</span><span class="mi">1</span><span class="p">,</span> <span class="n">device</span><span class="o">=</span><span class="n">device</span><span class="p">)</span>
        <span class="n">target_classes</span><span class="p">,</span> <span class="n">target_boxes</span><span class="p">,</span> <span class="n">indices</span><span class="p">,</span> <span class="n">anchors</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">build_targets</span><span class="p">(</span><span class="n">predictions</span><span class="p">,</span> <span class="n">targets</span><span class="p">)</span>  <span class="c1"># targets</span>

        <span class="c1"># Define criteria</span>
        <span class="n">BCEcls</span> <span class="o">=</span> <span class="n">nn</span><span class="o">.</span><span class="n">BCEWithLogitsLoss</span><span class="p">(</span><span class="n">pos_weight</span><span class="o">=</span><span class="n">torch</span><span class="o">.</span><span class="n">Tensor</span><span class="p">([</span><span class="bp">self</span><span class="o">.</span><span class="n">cls_pos_weight</span><span class="p">]))</span><span class="o">.</span><span class="n">to</span><span class="p">(</span><span class="n">device</span><span class="p">)</span>
        <span class="n">BCEobj</span> <span class="o">=</span> <span class="n">nn</span><span class="o">.</span><span class="n">BCEWithLogitsLoss</span><span class="p">(</span><span class="n">pos_weight</span><span class="o">=</span><span class="n">torch</span><span class="o">.</span><span class="n">Tensor</span><span class="p">([</span><span class="bp">self</span><span class="o">.</span><span class="n">obj_pos_weight</span><span class="p">]),</span> <span class="n">reduction</span><span class="o">=</span><span class="s1">&#39;none&#39;</span><span class="p">)</span><span class="o">.</span><span class="n">to</span><span class="p">(</span><span class="n">device</span><span class="p">)</span>

        <span class="c1"># Focal loss</span>
        <span class="k">if</span> <span class="bp">self</span><span class="o">.</span><span class="n">focal_loss_gamma</span> <span class="o">&gt;</span> <span class="mi">0</span><span class="p">:</span>
            <span class="n">BCEcls</span><span class="p">,</span> <span class="n">BCEobj</span> <span class="o">=</span> <span class="n">FocalLoss</span><span class="p">(</span><span class="n">BCEcls</span><span class="p">,</span> <span class="bp">self</span><span class="o">.</span><span class="n">focal_loss_gamma</span><span class="p">),</span> <span class="n">FocalLoss</span><span class="p">(</span><span class="n">BCEobj</span><span class="p">,</span> <span class="bp">self</span><span class="o">.</span><span class="n">focal_loss_gamma</span><span class="p">)</span>

        <span class="c1"># Losses</span>
        <span class="n">num_targets</span> <span class="o">=</span> <span class="mi">0</span>
        <span class="n">num_predictions</span> <span class="o">=</span> <span class="nb">len</span><span class="p">(</span><span class="n">predictions</span><span class="p">)</span>
        <span class="n">balance</span> <span class="o">=</span> <span class="p">[</span><span class="mf">4.0</span><span class="p">,</span> <span class="mf">1.0</span><span class="p">,</span> <span class="mf">0.4</span><span class="p">]</span> <span class="k">if</span> <span class="n">num_predictions</span> <span class="o">==</span> <span class="mi">3</span> <span class="k">else</span> <span class="p">[</span><span class="mf">4.0</span><span class="p">,</span> <span class="mf">1.0</span><span class="p">,</span> <span class="mf">0.4</span><span class="p">,</span> <span class="mf">0.1</span><span class="p">]</span>  <span class="c1"># P3-5 or P3-6</span>
        <span class="k">for</span> <span class="n">i</span><span class="p">,</span> <span class="n">prediction</span> <span class="ow">in</span> <span class="nb">enumerate</span><span class="p">(</span><span class="n">predictions</span><span class="p">):</span>  <span class="c1"># layer index, layer predictions</span>
            <span class="n">image</span><span class="p">,</span> <span class="n">anchor</span><span class="p">,</span> <span class="n">grid_y</span><span class="p">,</span> <span class="n">grid_x</span> <span class="o">=</span> <span class="n">indices</span><span class="p">[</span><span class="n">i</span><span class="p">]</span>
            <span class="n">target_obj</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">zeros_like</span><span class="p">(</span><span class="n">prediction</span><span class="p">[</span><span class="o">...</span><span class="p">,</span> <span class="mi">0</span><span class="p">],</span> <span class="n">device</span><span class="o">=</span><span class="n">device</span><span class="p">)</span>
            <span class="n">weight_obj</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">ones_like</span><span class="p">(</span><span class="n">prediction</span><span class="p">[</span><span class="o">...</span><span class="p">,</span> <span class="mi">0</span><span class="p">],</span> <span class="n">device</span><span class="o">=</span><span class="n">device</span><span class="p">)</span>

            <span class="n">n</span> <span class="o">=</span> <span class="n">image</span><span class="o">.</span><span class="n">shape</span><span class="p">[</span><span class="mi">0</span><span class="p">]</span>  <span class="c1"># number of targets</span>
            <span class="k">if</span> <span class="n">n</span><span class="p">:</span>
                <span class="n">num_targets</span> <span class="o">+=</span> <span class="n">n</span>  <span class="c1"># cumulative targets</span>
                <span class="n">ps</span> <span class="o">=</span> <span class="n">prediction</span><span class="p">[</span><span class="n">image</span><span class="p">,</span> <span class="n">anchor</span><span class="p">,</span> <span class="n">grid_y</span><span class="p">,</span> <span class="n">grid_x</span><span class="p">]</span>  <span class="c1"># prediction subset corresponding to targets</span>

                <span class="c1"># Boxes loss</span>
                <span class="n">pxy</span> <span class="o">=</span> <span class="n">ps</span><span class="p">[:,</span> <span class="p">:</span><span class="mi">2</span><span class="p">]</span><span class="o">.</span><span class="n">sigmoid</span><span class="p">()</span> <span class="o">*</span> <span class="mf">2.</span> <span class="o">-</span> <span class="mf">0.5</span>
                <span class="n">pwh</span> <span class="o">=</span> <span class="p">(</span><span class="n">ps</span><span class="p">[:,</span> <span class="mi">2</span><span class="p">:</span><span class="mi">4</span><span class="p">]</span><span class="o">.</span><span class="n">sigmoid</span><span class="p">()</span> <span class="o">*</span> <span class="mi">2</span><span class="p">)</span> <span class="o">**</span> <span class="mi">2</span> <span class="o">*</span> <span class="n">anchors</span><span class="p">[</span><span class="n">i</span><span class="p">]</span>
                <span class="n">pbox</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">cat</span><span class="p">((</span><span class="n">pxy</span><span class="p">,</span> <span class="n">pwh</span><span class="p">),</span> <span class="mi">1</span><span class="p">)</span><span class="o">.</span><span class="n">to</span><span class="p">(</span><span class="n">device</span><span class="p">)</span>  <span class="c1"># predicted box</span>
                <span class="n">iou</span> <span class="o">=</span> <span class="n">calculate_bbox_iou_elementwise</span><span class="p">(</span><span class="n">pbox</span><span class="o">.</span><span class="n">T</span><span class="p">,</span> <span class="n">target_boxes</span><span class="p">[</span><span class="n">i</span><span class="p">],</span> <span class="n">x1y1x2y2</span><span class="o">=</span><span class="kc">False</span><span class="p">,</span> <span class="n">CIoU</span><span class="o">=</span><span class="kc">True</span><span class="p">)</span>
                <span class="n">loss_boxes</span> <span class="o">+=</span> <span class="p">(</span><span class="mf">1.0</span> <span class="o">-</span> <span class="n">iou</span><span class="p">)</span><span class="o">.</span><span class="n">mean</span><span class="p">()</span>  <span class="c1"># iou loss</span>

                <span class="c1"># Objectness loss target</span>
                <span class="n">target_obj</span><span class="p">[</span><span class="n">image</span><span class="p">,</span> <span class="n">anchor</span><span class="p">,</span> <span class="n">grid_y</span><span class="p">,</span> <span class="n">grid_x</span><span class="p">]</span> <span class="o">=</span> \
                    <span class="p">(</span><span class="mf">1.0</span> <span class="o">-</span> <span class="n">giou_loss_ratio</span><span class="p">)</span> <span class="o">+</span> <span class="n">giou_loss_ratio</span> <span class="o">*</span> <span class="n">iou</span><span class="o">.</span><span class="n">detach</span><span class="p">()</span><span class="o">.</span><span class="n">clamp</span><span class="p">(</span><span class="mi">0</span><span class="p">)</span><span class="o">.</span><span class="n">type</span><span class="p">(</span><span class="n">target_obj</span><span class="o">.</span><span class="n">dtype</span><span class="p">)</span>
                <span class="c1"># Weights for weighted objectness</span>
                <span class="k">if</span> <span class="bp">self</span><span class="o">.</span><span class="n">cls_obj_weights</span> <span class="ow">is</span> <span class="ow">not</span> <span class="kc">None</span><span class="p">:</span>
                    <span class="c1"># NOTE: for grid cells that have a few ground truths with different classes assigned to them</span>
                    <span class="c1"># objectness weight will be picked randomly from one of these classes</span>
                    <span class="n">weight_obj</span><span class="p">[</span><span class="n">image</span><span class="p">,</span> <span class="n">anchor</span><span class="p">,</span> <span class="n">grid_y</span><span class="p">,</span> <span class="n">grid_x</span><span class="p">]</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">cls_obj_weights</span><span class="p">[</span><span class="n">target_classes</span><span class="p">[</span><span class="n">i</span><span class="p">]]</span>

                <span class="c1"># Classification loss</span>
                <span class="k">if</span> <span class="n">ps</span><span class="o">.</span><span class="n">shape</span><span class="p">[</span><span class="mi">1</span><span class="p">]</span> <span class="o">&gt;</span> <span class="mi">6</span><span class="p">:</span>   <span class="c1"># cls loss (only if multiple classes)</span>
                    <span class="n">t</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">full_like</span><span class="p">(</span><span class="n">ps</span><span class="p">[:,</span> <span class="mi">5</span><span class="p">:],</span> <span class="mi">0</span><span class="p">,</span> <span class="n">device</span><span class="o">=</span><span class="n">device</span><span class="p">)</span>  <span class="c1"># targets</span>
                    <span class="n">t</span><span class="p">[</span><span class="nb">range</span><span class="p">(</span><span class="n">n</span><span class="p">),</span> <span class="n">target_classes</span><span class="p">[</span><span class="n">i</span><span class="p">]]</span> <span class="o">=</span> <span class="mi">1</span>
                    <span class="n">loss_classification</span> <span class="o">+=</span> <span class="n">BCEcls</span><span class="p">(</span><span class="n">ps</span><span class="p">[:,</span> <span class="mi">5</span><span class="p">:],</span> <span class="n">t</span><span class="p">)</span>  <span class="c1"># BCE</span>

            <span class="c1"># Objectness loss</span>
            <span class="n">loss_obj_cur_head</span> <span class="o">=</span> <span class="n">BCEobj</span><span class="p">(</span><span class="n">prediction</span><span class="p">[</span><span class="o">...</span><span class="p">,</span> <span class="mi">4</span><span class="p">],</span> <span class="n">target_obj</span><span class="p">)</span>
            <span class="n">loss_obj_cur_head</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">sum</span><span class="p">(</span><span class="n">loss_obj_cur_head</span> <span class="o">*</span> <span class="n">weight_obj</span> <span class="o">/</span> <span class="n">torch</span><span class="o">.</span><span class="n">sum</span><span class="p">(</span><span class="n">weight_obj</span><span class="p">))</span>
            <span class="n">loss_objectivness</span> <span class="o">+=</span> <span class="n">loss_obj_cur_head</span> <span class="o">*</span> <span class="n">balance</span><span class="p">[</span><span class="n">i</span><span class="p">]</span>  <span class="c1"># obj loss</span>

        <span class="n">batch_size</span> <span class="o">=</span> <span class="n">prediction</span><span class="o">.</span><span class="n">shape</span><span class="p">[</span><span class="mi">0</span><span class="p">]</span>  <span class="c1"># batch size</span>

        <span class="n">loss</span> <span class="o">=</span> <span class="n">loss_boxes</span> <span class="o">*</span> <span class="bp">self</span><span class="o">.</span><span class="n">box_loss_gain</span> <span class="o">+</span> <span class="n">loss_objectivness</span> <span class="o">*</span> <span class="bp">self</span><span class="o">.</span><span class="n">obj_loss_gain</span> <span class="o">+</span> <span class="n">loss_classification</span> <span class="o">*</span> <span class="bp">self</span><span class="o">.</span><span class="n">cls_loss_gain</span>
        <span class="c1"># IMPORTANT: box, obj and cls loss are logged scaled by gain in ultralytics</span>
        <span class="c1"># and are logged unscaled in our codebase</span>
        <span class="k">return</span> <span class="n">loss</span> <span class="o">*</span> <span class="n">batch_size</span><span class="p">,</span> <span class="n">torch</span><span class="o">.</span><span class="n">cat</span><span class="p">((</span><span class="n">loss_boxes</span><span class="p">,</span> <span class="n">loss_objectivness</span><span class="p">,</span> <span class="n">loss_classification</span><span class="p">,</span> <span class="n">loss</span><span class="p">))</span><span class="o">.</span><span class="n">detach</span><span class="p">()</span></div></div>
</pre></div>

           </div>
          </div>
          <footer>

  <hr/>

  <div role="contentinfo">
    <p>&#169; Copyright 2021, SuperGradients team.</p>
  </div>

  Built with <a href="https://www.sphinx-doc.org/">Sphinx</a> using a
    <a href="https://github.com/readthedocs/sphinx_rtd_theme">theme</a>
    provided by <a href="https://readthedocs.org">Read the Docs</a>.
   

</footer>
        </div>
      </div>
    </section>
  </div>
  <script>
      jQuery(function () {
          SphinxRtdTheme.Navigation.enable(true);
      });
  </script> 

</body>
</html>