yzbx
commented
6 years ago psp_convert -dropout -relu -batchnorm
----------------------------------------------------------------
Layer (type) Output Shape Param #
================================================================
Conv2d-1 [-1, 64, 357, 357] 1,728
Conv2d-4 [-1, 64, 357, 357] 36,864
Conv2d-7 [-1, 128, 357, 357] 73,728
MaxPool2d-10 [-1, 128, 179, 179] 0
Conv2d-11 [-1, 64, 179, 179] 8,192
Conv2d-14 [-1, 64, 179, 179] 36,864
Conv2d-17 [-1, 256, 179, 179] 16,384
Conv2d-19 [-1, 256, 179, 179] 32,768
Bottleneck-22 [-1, 256, 179, 179] 0
Conv2d-23 [-1, 64, 179, 179] 16,384
Conv2d-26 [-1, 64, 179, 179] 36,864
Conv2d-29 [-1, 256, 179, 179] 16,384
Bottleneck-32 [-1, 256, 179, 179] 0
Conv2d-33 [-1, 64, 179, 179] 16,384
Conv2d-36 [-1, 64, 179, 179] 36,864
Conv2d-39 [-1, 256, 179, 179] 16,384
Bottleneck-42 [-1, 256, 179, 179] 0
Conv2d-43 [-1, 128, 179, 179] 32,768
Conv2d-46 [-1, 128, 90, 90] 147,456
Conv2d-49 [-1, 512, 90, 90] 65,536
Conv2d-51 [-1, 512, 90, 90] 131,072
Bottleneck-54 [-1, 512, 90, 90] 0
Conv2d-55 [-1, 128, 90, 90] 65,536
Conv2d-58 [-1, 128, 90, 90] 147,456
Conv2d-61 [-1, 512, 90, 90] 65,536
Bottleneck-64 [-1, 512, 90, 90] 0
Conv2d-65 [-1, 128, 90, 90] 65,536
Conv2d-68 [-1, 128, 90, 90] 147,456
Conv2d-71 [-1, 512, 90, 90] 65,536
Bottleneck-74 [-1, 512, 90, 90] 0
Conv2d-75 [-1, 128, 90, 90] 65,536
Conv2d-78 [-1, 128, 90, 90] 147,456
Conv2d-81 [-1, 512, 90, 90] 65,536
Bottleneck-84 [-1, 512, 90, 90] 0
Conv2d-85 [-1, 256, 90, 90] 131,072
Conv2d-88 [-1, 256, 90, 90] 589,824
Conv2d-91 [-1, 1024, 90, 90] 262,144
Conv2d-93 [-1, 1024, 90, 90] 524,288
Bottleneck-96 [-1, 1024, 90, 90] 0
Conv2d-97 [-1, 256, 90, 90] 262,144
Conv2d-100 [-1, 256, 90, 90] 589,824
Conv2d-103 [-1, 1024, 90, 90] 262,144
Bottleneck-106 [-1, 1024, 90, 90] 0
Conv2d-107 [-1, 256, 90, 90] 262,144
Conv2d-110 [-1, 256, 90, 90] 589,824
Conv2d-113 [-1, 1024, 90, 90] 262,144
Bottleneck-116 [-1, 1024, 90, 90] 0
Conv2d-117 [-1, 256, 90, 90] 262,144
Conv2d-120 [-1, 256, 90, 90] 589,824
Conv2d-123 [-1, 1024, 90, 90] 262,144
Bottleneck-126 [-1, 1024, 90, 90] 0
Conv2d-127 [-1, 256, 90, 90] 262,144
Conv2d-130 [-1, 256, 90, 90] 589,824
Conv2d-133 [-1, 1024, 90, 90] 262,144
Bottleneck-136 [-1, 1024, 90, 90] 0
Conv2d-137 [-1, 256, 90, 90] 262,144
Conv2d-140 [-1, 256, 90, 90] 589,824
Conv2d-143 [-1, 1024, 90, 90] 262,144
Bottleneck-146 [-1, 1024, 90, 90] 0
Conv2d-147 [-1, 256, 90, 90] 262,144
Conv2d-150 [-1, 256, 90, 90] 589,824
Conv2d-153 [-1, 1024, 90, 90] 262,144
Bottleneck-156 [-1, 1024, 90, 90] 0
Conv2d-157 [-1, 256, 90, 90] 262,144
Conv2d-160 [-1, 256, 90, 90] 589,824
Conv2d-163 [-1, 1024, 90, 90] 262,144
Bottleneck-166 [-1, 1024, 90, 90] 0
Conv2d-167 [-1, 256, 90, 90] 262,144
Conv2d-170 [-1, 256, 90, 90] 589,824
Conv2d-173 [-1, 1024, 90, 90] 262,144
Bottleneck-176 [-1, 1024, 90, 90] 0
Conv2d-177 [-1, 256, 90, 90] 262,144
Conv2d-180 [-1, 256, 90, 90] 589,824
Conv2d-183 [-1, 1024, 90, 90] 262,144
Bottleneck-186 [-1, 1024, 90, 90] 0
Conv2d-187 [-1, 256, 90, 90] 262,144
Conv2d-190 [-1, 256, 90, 90] 589,824
Conv2d-193 [-1, 1024, 90, 90] 262,144
Bottleneck-196 [-1, 1024, 90, 90] 0
Conv2d-197 [-1, 256, 90, 90] 262,144
Conv2d-200 [-1, 256, 90, 90] 589,824
Conv2d-203 [-1, 1024, 90, 90] 262,144
Bottleneck-206 [-1, 1024, 90, 90] 0
Conv2d-207 [-1, 256, 90, 90] 262,144
Conv2d-210 [-1, 256, 90, 90] 589,824
Conv2d-213 [-1, 1024, 90, 90] 262,144
Bottleneck-216 [-1, 1024, 90, 90] 0
Conv2d-217 [-1, 256, 90, 90] 262,144
Conv2d-220 [-1, 256, 90, 90] 589,824
Conv2d-223 [-1, 1024, 90, 90] 262,144
Bottleneck-226 [-1, 1024, 90, 90] 0
Conv2d-227 [-1, 256, 90, 90] 262,144
Conv2d-230 [-1, 256, 90, 90] 589,824
Conv2d-233 [-1, 1024, 90, 90] 262,144
Bottleneck-236 [-1, 1024, 90, 90] 0
Conv2d-237 [-1, 256, 90, 90] 262,144
Conv2d-240 [-1, 256, 90, 90] 589,824
Conv2d-243 [-1, 1024, 90, 90] 262,144
Bottleneck-246 [-1, 1024, 90, 90] 0
Conv2d-247 [-1, 256, 90, 90] 262,144
Conv2d-250 [-1, 256, 90, 90] 589,824
Conv2d-253 [-1, 1024, 90, 90] 262,144
Bottleneck-256 [-1, 1024, 90, 90] 0
Conv2d-257 [-1, 256, 90, 90] 262,144
Conv2d-260 [-1, 256, 90, 90] 589,824
Conv2d-263 [-1, 1024, 90, 90] 262,144
Bottleneck-266 [-1, 1024, 90, 90] 0
Conv2d-267 [-1, 256, 90, 90] 262,144
Conv2d-270 [-1, 256, 90, 90] 589,824
Conv2d-273 [-1, 1024, 90, 90] 262,144
Bottleneck-276 [-1, 1024, 90, 90] 0
Conv2d-277 [-1, 256, 90, 90] 262,144
Conv2d-280 [-1, 256, 90, 90] 589,824
Conv2d-283 [-1, 1024, 90, 90] 262,144
Bottleneck-286 [-1, 1024, 90, 90] 0
Conv2d-287 [-1, 256, 90, 90] 262,144
Conv2d-290 [-1, 256, 90, 90] 589,824
Conv2d-293 [-1, 1024, 90, 90] 262,144
Bottleneck-296 [-1, 1024, 90, 90] 0
Conv2d-297 [-1, 256, 90, 90] 262,144
Conv2d-300 [-1, 256, 90, 90] 589,824
Conv2d-303 [-1, 1024, 90, 90] 262,144
Bottleneck-306 [-1, 1024, 90, 90] 0
Conv2d-307 [-1, 256, 90, 90] 262,144
Conv2d-310 [-1, 256, 90, 90] 589,824
Conv2d-313 [-1, 1024, 90, 90] 262,144
Bottleneck-316 [-1, 1024, 90, 90] 0
Conv2d-317 [-1, 256, 90, 90] 2,359,296
conv_bn_relu-320 [-1, 256, 90, 90] 0
Conv2d-322 [-1, 19, 90, 90] 4,883
Conv2d-323 [-1, 512, 90, 90] 524,288
Conv2d-326 [-1, 512, 90, 90] 2,359,296
Conv2d-329 [-1, 2048, 90, 90] 1,048,576
Conv2d-331 [-1, 2048, 90, 90] 2,097,152
Bottleneck-334 [-1, 2048, 90, 90] 0
Conv2d-335 [-1, 512, 90, 90] 1,048,576
Conv2d-338 [-1, 512, 90, 90] 2,359,296
Conv2d-341 [-1, 2048, 90, 90] 1,048,576
Bottleneck-344 [-1, 2048, 90, 90] 0
Conv2d-345 [-1, 512, 90, 90] 1,048,576
Conv2d-348 [-1, 512, 90, 90] 2,359,296
Conv2d-351 [-1, 2048, 90, 90] 1,048,576
Bottleneck-354 [-1, 2048, 90, 90] 0
AvgPool2d-355 [-1, 2048, 6, 6] 0
Conv2d-356 [-1, 512, 6, 6] 1,048,576
Upsample-359 [-1, 512, 90, 90] 0
AvgPool2d-360 [-1, 2048, 3, 3] 0
Conv2d-361 [-1, 512, 3, 3] 1,048,576
Upsample-364 [-1, 512, 90, 90] 0
AvgPool2d-365 [-1, 2048, 2, 2] 0
Conv2d-366 [-1, 512, 2, 2] 1,048,576
Upsample-369 [-1, 512, 90, 90] 0
AvgPool2d-370 [-1, 2048, 1, 1] 0
Conv2d-371 [-1, 512, 1, 1] 1,048,576
Upsample-374 [-1, 512, 90, 90] 0
transform_psp_caffe-375 [-1, 4096, 90, 90] 0
Conv2d-376 [-1, 512, 90, 90] 18,874,368
conv_bn_relu-379 [-1, 512, 90, 90] 0
Conv2d-381 [-1, 19, 90, 90] 9,747
Upsample-382 [-1, 19, 713, 713] 0
================================================================
Total params: 68,072,166
Trainable params: 68,072,166
Non-trainable params: 0
----------------------------------------------------------------
pytorch
pytorch semantic segmentation weights