TensorRT-Plugin实现之reorgPlugin算子实现-cuda-TRT8.zip

# reorgPlugin **Table Of Contents** - [Description](#description) * [Structure](#structure) - [Parameters](#parameters) - [Additional resources](#additional-resources) - [License](#license) - [Changelog](#changelog) - [Known issues](#known-issues) ## Description The`reorgPlugin` is specifically used for the reorg layer in the YOLOv2 model in TensorRT. It reorganizes the elements in the input tensor and generates an output tensor of a different shape. In YOLOv2, the output tensor from the reorg layer matches the shape of the output tensor from a downstream layer Conv20_1024 in the neural network. The two output tensors are then concatenated together as one single tensor. ### Structure The `reorgPlugin` takes one input and generates one output. The tensor format must be in NCHW format. The input is a tensor that has a shape of `[N, C, H, W]` where: - `N` is the batch size - `C` is the number of channels - `H` is the height of tensor - `W` is the width of the tensor After a [unique one-to-one mapping](https://github.com/pjreddie/darknet/blob/8215a8864d4ad07e058acafd75b2c6ff6600b9e8/src/blas.c#L9), the output tensor of shape `[N, C x s x s, H / s, W / s]`, where s is the stride, is generated. For example, if we have an input tensor of shape `[2, 4, 6, 6]`. ``` [[[[ 0 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]]]] ``` We set `stride = 2` and perform the reorganization, we will get the following output tensor of shape `[2, 16, 3, 3]`. ``` [[[[ 0 2 4] [ 6 8 10] [ 24 26 28]] [[ 30 32 34] [ 48 50 52] [ 54 56 58]] [[ 72 74 76] [ 78 80 82] [ 96 98 100]] [[102 104 106] [120 122 124] [126 128 130]] [[ 1 3 5] [ 7 9 11] [ 25 27 29]] [[ 31 33 35] [ 49 51 53] [ 55 57 59]] [[ 73 75 77] [ 79 81 83] [ 97 99 101]] [[103 105 107] [121 123 125] [127 129 131]] [[ 12 14 16] [ 18 20 22] [ 36 38 40]] [[ 42 44 46] [ 60 62 64] [ 66 68 70]] [[ 84 86 88] [ 90 92 94] [108 110 112]] [[114 116 118] [132 134 136] [138 140 142]] [[ 13 15 17] [ 19 21 23] [ 37 39 41]] [[ 43 45 47] [ 61 63 65] [ 67 69 71]] [[ 85 87 89] [ 91 93 95] [109 111 113]] [[115 117 119] [133 135 137] [139 141 143]]] [[[144 146 148] [150 152 154] [168 170 172]] [[174 176 178] [192 194 196] [198 200 202]] [[216 218 220] [222 224 226] [240 242 244]] [[246 248 250] [264 266 268] [270 272 274]] [[145 147 149] [151 153 155] [169 171 173]] [[175 177 179] [193 195 197] [199 201 203]] [[217 219 221] [223 225 227] [241 243 245]] [[247 249 251] [265 267 269] [271 273 275]] [[156 158 160] [162 164 166] [180 182 184]] [[186 188 190] [204 206 208] [210 212 214]] [[228 230 232] [234 236 238] [252 254 256]] [[258 260 262] [276 278 280] [282 284 286]] [[157 159 161] [163 165 167] [181 183 185]] [[187 189 191] [205 207 209] [211 213 215]] [[229 231 233] [235 237 239] [253 255 257]] [[259 261 263] [277 279 281] [283 285 287]]]] ``` ## Parameters The `reorgPlugin` has plugin creator class `ReorgPluginCreator` and plugin class `Reorg`. The following parameters were used to create the `Reorg` instance. | Type | Parameter | Description |----------|--------------------------|-------------------------------------------------------- |`int` |`stride` |Dimension reduction factor for the input tensor. The stride has to be divisible by the height and the width of the input tensor. ## Additional resources The following resources provide a deeper understanding of the `regionPlugin` plugin: - [YOLOv2 paper](https://arxiv.org/abs/1612.08242) - [Reorg layer in YOLOv2](https://github.com/pjreddie/darknet/blob/8215a8864d4ad07e058acafd75b2c6ff6600b9e8/src/blas.c#L9) - [YOLOv2 architecture](https://ethereon.github.io/netscope/#/gist/d08a41711e48cf111e330827b1279c31)