├── LICENSE
├── README.md
├── SqueezeNet_v1.0
    ├── deploy.prototxt
    ├── solver.prototxt
    ├── squeezenet_v1.0.caffemodel
    └── train_val.prototxt
└── SqueezeNet_v1.1
    ├── README.md
    ├── deploy.prototxt
    ├── solver.prototxt
    ├── squeezenet_v1.1.caffemodel
    └── train_val.prototxt


/LICENSE:
--------------------------------------------------------------------------------
 1 | BSD LICENSE.
 2 | 
 3 | Redistribution and use in source and binary forms, with or without modification, are permitted
 4 | provided that the following conditions are met:
 5 | 
 6 | 1. Redistributions of source code must retain the above copyright notice, this list of conditions
 7 | and the following disclaimer.
 8 | 
 9 | 2. Redistributions in binary form must reproduce the above copyright notice, this list of conditions
10 | and the following disclaimer in the documentation and/or other materials provided with the
11 | distribution.
12 | 
13 | THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR
14 | IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND
15 | FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR
16 | CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
17 | DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
18 | DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER
19 | IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF
20 | THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
21 | 


--------------------------------------------------------------------------------
/README.md:
--------------------------------------------------------------------------------
 1 | 
 2 | The Caffe-compatible files that you are probably looking for:
 3 | 
 4 |     SqueezeNet_v1.0/train_val.prototxt          #model architecture
 5 |     SqueezeNet_v1.0/solver.prototxt             #additional training details (learning rate schedule, etc.)
 6 |     SqueezeNet_v1.0/squeezenet_v1.0.caffemodel  #pretrained model parameters
 7 | 
 8 | If you find SqueezeNet useful in your research, please consider citing the [SqueezeNet paper](http://arxiv.org/abs/1602.07360):
 9 | 
10 |     @article{SqueezeNet,
11 |         Author = {Forrest N. Iandola and Song Han and Matthew W. Moskewicz and Khalid Ashraf and William J. Dally and Kurt Keutzer},
12 |         Title = {SqueezeNet: AlexNet-level accuracy with 50x fewer parameters and $<$0.5MB model size},
13 |         Journal = {arXiv:1602.07360},
14 |         Year = {2016}
15 |     }
16 | 
17 | 
18 | Helpful hints:
19 | 
20 | 1. **Getting the SqueezeNet model:** `git clone <this repo>`.
21 | In this repository, we include Caffe-compatible files for the model architecture, the solver configuration, and the pretrained model (4.8MB uncompressed).
22 | 
23 | 2. **Batch size.** We have experimented with batch sizes ranging from 32 to 1024. In this repo, our default batch size is 512. If implemented naively on a single GPU, a batch size this large may result in running out of memory. An effective workaround is to use hierarchical batching (sometimes called "delayed batching"). Caffe supports hierarchical batching by doing `train_val.prototxt>batch_size` training samples concurrently in memory. After `solver.prototxt>iter_size` iterations, the gradients are summed and the model is updated. Mathematically, the batch size is `batch_size * iter_size`. In the included prototxt files, we have set `(batch_size=32, iter_size=16)`, but any combination of batch_size and iter_size that multiply to 512 will produce eqivalent results. In fact, with the same random number generator seed, the model will be fully reproducable if trained multiple times. Finally, note that in Caffe `iter_size` is applied while training on the training set but not while testing on  the test set.
24 | 
25 | 3. **Implementing Fire modules.** In the paper, we describe the `expand` portion of the Fire layer as a collection of 1x1 and 3x3 filters. Caffe does not natively support a convolution layer that has multiple filter sizes. To work around this, we implement `expand1x1` and `expand3x3` layers and concatenate the results together in the channel dimension.
26 | 
27 | 4. **The SqueezeNet team has released a few variants of SqueezeNet**. Each of these include pretrained models, and the non-compressed versions include training protocols, too.
28 | 
29 |   SqueezeNet v1.0 (in this repo), the base model described in our SqueezeNet paper.
30 | 
31 |   [Compressed SqueezeNet v1.0](https://github.com/songhan/SqueezeNet_compressed), as described in the SqueezeNet paper.
32 | 
33 |   [SqueezeNet v1.0 with Residual Connections](https://github.com/songhan/SqueezeNet-Residual), which delivers higher accuracy without increasing the model size.
34 | 
35 |   [SqueezeNet v1.0 with Dense→Sparse→Dense (DSD) Training](https://github.com/songhan/SqueezeNet-DSD-Training), which delivers higher accuracy without increasing the model size.
36 | 
37 |   SqueezeNet v1.1 (in this repo), which requires 2.4x less computation than SqueezeNet v1.0 without diminshing accuracy.
38 | 
39 | 5. **Community adoption of SqueezeNet**:
40 | 
41 |   [SqueezeNet in the *MXNet* framework](https://github.com/haria/SqueezeNet), by Guo Haria
42 | 
43 |   [SqueezeNet in the *Chainer* framework](https://github.com/ejlb/squeezenet-chainer), by Eddie Bell
44 | 
45 |   [SqueezeNet in the *Keras* framework](https://github.com/DT42/squeezenet_demo), by [dt42.io](https://dt42.io/)
46 |   
47 |   [SqueezeNet in the *Tensorflow* framework](https://github.com/vonclites/squeezenet), by Domenick Poster
48 | 
49 |   [SqueezeNet in the *PyTorch* framework](https://github.com/pytorch/vision/blob/master/torchvision/models/squeezenet.py), by Marat Dukhan
50 | 
51 |   [SqueezeNet in the *CoreML* framework](https://github.com/mdering/CoreMLZoo)
52 | 
53 |   [Neural Art using SqueezeNet](https://github.com/pavelgonchar/neural-art-mini), by Pavel Gonchar
54 | 
55 |   [SqueezeNet compression in Ristretto](https://arxiv.org/abs/1605.06402), by Philipp Gysel
56 |   
57 | **If you like SqueezeNet, you might also like SqueezeNext! ([SqueezeNext paper](https://arxiv.org/abs/1803.10615), [SqueezeNext code](https://github.com/amirgholami/SqueezeNext))**
58 | 


--------------------------------------------------------------------------------
/SqueezeNet_v1.0/deploy.prototxt:
--------------------------------------------------------------------------------
  1 | # please cite:
  2 | # @article{SqueezeNet,
  3 | #     Author = {Forrest N. Iandola and Matthew W. Moskewicz and Khalid Ashraf and Song Han and William J. Dally and Kurt Keutzer},
  4 | #     Title = {SqueezeNet: AlexNet-level accuracy with 50x fewer parameters and $<$1MB model size},
  5 | #     Journal = {arXiv:1602.07360},
  6 | #     Year = {2016}
  7 | #}
  8 | 
  9 | input: "data"
 10 | input_shape {
 11 |   dim: 10
 12 |   dim: 3
 13 |   dim: 227
 14 |   dim: 227
 15 | }
 16 | layer {
 17 |   name: "conv1"
 18 |   type: "Convolution"
 19 |   bottom: "data"
 20 |   top: "conv1"
 21 |   convolution_param {
 22 |     num_output: 96
 23 |     kernel_size: 7
 24 |     stride: 2
 25 |   }
 26 | }
 27 | layer {
 28 |   name: "relu_conv1"
 29 |   type: "ReLU"
 30 |   bottom: "conv1"
 31 |   top: "conv1"
 32 | }
 33 | layer {
 34 |   name: "pool1"
 35 |   type: "Pooling"
 36 |   bottom: "conv1"
 37 |   top: "pool1"
 38 |   pooling_param {
 39 |     pool: MAX
 40 |     kernel_size: 3
 41 |     stride: 2
 42 |   }
 43 | }
 44 | layer {
 45 |   name: "fire2/squeeze1x1"
 46 |   type: "Convolution"
 47 |   bottom: "pool1"
 48 |   top: "fire2/squeeze1x1"
 49 |   convolution_param {
 50 |     num_output: 16
 51 |     kernel_size: 1
 52 |   }
 53 | }
 54 | layer {
 55 |   name: "fire2/relu_squeeze1x1"
 56 |   type: "ReLU"
 57 |   bottom: "fire2/squeeze1x1"
 58 |   top: "fire2/squeeze1x1"
 59 | }
 60 | layer {
 61 |   name: "fire2/expand1x1"
 62 |   type: "Convolution"
 63 |   bottom: "fire2/squeeze1x1"
 64 |   top: "fire2/expand1x1"
 65 |   convolution_param {
 66 |     num_output: 64
 67 |     kernel_size: 1
 68 |   }
 69 | }
 70 | layer {
 71 |   name: "fire2/relu_expand1x1"
 72 |   type: "ReLU"
 73 |   bottom: "fire2/expand1x1"
 74 |   top: "fire2/expand1x1"
 75 | }
 76 | layer {
 77 |   name: "fire2/expand3x3"
 78 |   type: "Convolution"
 79 |   bottom: "fire2/squeeze1x1"
 80 |   top: "fire2/expand3x3"
 81 |   convolution_param {
 82 |     num_output: 64
 83 |     pad: 1
 84 |     kernel_size: 3
 85 |   }
 86 | }
 87 | layer {
 88 |   name: "fire2/relu_expand3x3"
 89 |   type: "ReLU"
 90 |   bottom: "fire2/expand3x3"
 91 |   top: "fire2/expand3x3"
 92 | }
 93 | layer {
 94 |   name: "fire2/concat"
 95 |   type: "Concat"
 96 |   bottom: "fire2/expand1x1"
 97 |   bottom: "fire2/expand3x3"
 98 |   top: "fire2/concat"
 99 | }
100 | layer {
101 |   name: "fire3/squeeze1x1"
102 |   type: "Convolution"
103 |   bottom: "fire2/concat"
104 |   top: "fire3/squeeze1x1"
105 |   convolution_param {
106 |     num_output: 16
107 |     kernel_size: 1
108 |   }
109 | }
110 | layer {
111 |   name: "fire3/relu_squeeze1x1"
112 |   type: "ReLU"
113 |   bottom: "fire3/squeeze1x1"
114 |   top: "fire3/squeeze1x1"
115 | }
116 | layer {
117 |   name: "fire3/expand1x1"
118 |   type: "Convolution"
119 |   bottom: "fire3/squeeze1x1"
120 |   top: "fire3/expand1x1"
121 |   convolution_param {
122 |     num_output: 64
123 |     kernel_size: 1
124 |   }
125 | }
126 | layer {
127 |   name: "fire3/relu_expand1x1"
128 |   type: "ReLU"
129 |   bottom: "fire3/expand1x1"
130 |   top: "fire3/expand1x1"
131 | }
132 | layer {
133 |   name: "fire3/expand3x3"
134 |   type: "Convolution"
135 |   bottom: "fire3/squeeze1x1"
136 |   top: "fire3/expand3x3"
137 |   convolution_param {
138 |     num_output: 64
139 |     pad: 1
140 |     kernel_size: 3
141 |   }
142 | }
143 | layer {
144 |   name: "fire3/relu_expand3x3"
145 |   type: "ReLU"
146 |   bottom: "fire3/expand3x3"
147 |   top: "fire3/expand3x3"
148 | }
149 | layer {
150 |   name: "fire3/concat"
151 |   type: "Concat"
152 |   bottom: "fire3/expand1x1"
153 |   bottom: "fire3/expand3x3"
154 |   top: "fire3/concat"
155 | }
156 | layer {
157 |   name: "fire4/squeeze1x1"
158 |   type: "Convolution"
159 |   bottom: "fire3/concat"
160 |   top: "fire4/squeeze1x1"
161 |   convolution_param {
162 |     num_output: 32
163 |     kernel_size: 1
164 |   }
165 | }
166 | layer {
167 |   name: "fire4/relu_squeeze1x1"
168 |   type: "ReLU"
169 |   bottom: "fire4/squeeze1x1"
170 |   top: "fire4/squeeze1x1"
171 | }
172 | layer {
173 |   name: "fire4/expand1x1"
174 |   type: "Convolution"
175 |   bottom: "fire4/squeeze1x1"
176 |   top: "fire4/expand1x1"
177 |   convolution_param {
178 |     num_output: 128
179 |     kernel_size: 1
180 |   }
181 | }
182 | layer {
183 |   name: "fire4/relu_expand1x1"
184 |   type: "ReLU"
185 |   bottom: "fire4/expand1x1"
186 |   top: "fire4/expand1x1"
187 | }
188 | layer {
189 |   name: "fire4/expand3x3"
190 |   type: "Convolution"
191 |   bottom: "fire4/squeeze1x1"
192 |   top: "fire4/expand3x3"
193 |   convolution_param {
194 |     num_output: 128
195 |     pad: 1
196 |     kernel_size: 3
197 |   }
198 | }
199 | layer {
200 |   name: "fire4/relu_expand3x3"
201 |   type: "ReLU"
202 |   bottom: "fire4/expand3x3"
203 |   top: "fire4/expand3x3"
204 | }
205 | layer {
206 |   name: "fire4/concat"
207 |   type: "Concat"
208 |   bottom: "fire4/expand1x1"
209 |   bottom: "fire4/expand3x3"
210 |   top: "fire4/concat"
211 | }
212 | layer {
213 |   name: "pool4"
214 |   type: "Pooling"
215 |   bottom: "fire4/concat"
216 |   top: "pool4"
217 |   pooling_param {
218 |     pool: MAX
219 |     kernel_size: 3
220 |     stride: 2
221 |  }
222 | }
223 | layer {
224 |   name: "fire5/squeeze1x1"
225 |   type: "Convolution"
226 |   bottom: "pool4"
227 |   top: "fire5/squeeze1x1"
228 |   convolution_param {
229 |     num_output: 32
230 |     kernel_size: 1
231 |   }
232 | }
233 | layer {
234 |   name: "fire5/relu_squeeze1x1"
235 |   type: "ReLU"
236 |   bottom: "fire5/squeeze1x1"
237 |   top: "fire5/squeeze1x1"
238 | }
239 | layer {
240 |   name: "fire5/expand1x1"
241 |   type: "Convolution"
242 |   bottom: "fire5/squeeze1x1"
243 |   top: "fire5/expand1x1"
244 |   convolution_param {
245 |     num_output: 128
246 |     kernel_size: 1
247 |   }
248 | }
249 | layer {
250 |   name: "fire5/relu_expand1x1"
251 |   type: "ReLU"
252 |   bottom: "fire5/expand1x1"
253 |   top: "fire5/expand1x1"
254 | }
255 | layer {
256 |   name: "fire5/expand3x3"
257 |   type: "Convolution"
258 |   bottom: "fire5/squeeze1x1"
259 |   top: "fire5/expand3x3"
260 |   convolution_param {
261 |     num_output: 128
262 |     pad: 1
263 |     kernel_size: 3
264 |   }
265 | }
266 | layer {
267 |   name: "fire5/relu_expand3x3"
268 |   type: "ReLU"
269 |   bottom: "fire5/expand3x3"
270 |   top: "fire5/expand3x3"
271 | }
272 | layer {
273 |   name: "fire5/concat"
274 |   type: "Concat"
275 |   bottom: "fire5/expand1x1"
276 |   bottom: "fire5/expand3x3"
277 |   top: "fire5/concat"
278 | }
279 | layer {
280 |   name: "fire6/squeeze1x1"
281 |   type: "Convolution"
282 |   bottom: "fire5/concat"
283 |   top: "fire6/squeeze1x1"
284 |   convolution_param {
285 |     num_output: 48
286 |     kernel_size: 1
287 |   }
288 | }
289 | layer {
290 |   name: "fire6/relu_squeeze1x1"
291 |   type: "ReLU"
292 |   bottom: "fire6/squeeze1x1"
293 |   top: "fire6/squeeze1x1"
294 | }
295 | layer {
296 |   name: "fire6/expand1x1"
297 |   type: "Convolution"
298 |   bottom: "fire6/squeeze1x1"
299 |   top: "fire6/expand1x1"
300 |   convolution_param {
301 |     num_output: 192
302 |     kernel_size: 1
303 |   }
304 | }
305 | layer {
306 |   name: "fire6/relu_expand1x1"
307 |   type: "ReLU"
308 |   bottom: "fire6/expand1x1"
309 |   top: "fire6/expand1x1"
310 | }
311 | layer {
312 |   name: "fire6/expand3x3"
313 |   type: "Convolution"
314 |   bottom: "fire6/squeeze1x1"
315 |   top: "fire6/expand3x3"
316 |   convolution_param {
317 |     num_output: 192
318 |     pad: 1
319 |     kernel_size: 3
320 |   }
321 | }
322 | layer {
323 |   name: "fire6/relu_expand3x3"
324 |   type: "ReLU"
325 |   bottom: "fire6/expand3x3"
326 |   top: "fire6/expand3x3"
327 | }
328 | layer {
329 |   name: "fire6/concat"
330 |   type: "Concat"
331 |   bottom: "fire6/expand1x1"
332 |   bottom: "fire6/expand3x3"
333 |   top: "fire6/concat"
334 | }
335 | layer {
336 |   name: "fire7/squeeze1x1"
337 |   type: "Convolution"
338 |   bottom: "fire6/concat"
339 |   top: "fire7/squeeze1x1"
340 |   convolution_param {
341 |     num_output: 48
342 |     kernel_size: 1
343 |   }
344 | }
345 | layer {
346 |   name: "fire7/relu_squeeze1x1"
347 |   type: "ReLU"
348 |   bottom: "fire7/squeeze1x1"
349 |   top: "fire7/squeeze1x1"
350 | }
351 | layer {
352 |   name: "fire7/expand1x1"
353 |   type: "Convolution"
354 |   bottom: "fire7/squeeze1x1"
355 |   top: "fire7/expand1x1"
356 |   convolution_param {
357 |     num_output: 192
358 |     kernel_size: 1
359 |   }
360 | }
361 | layer {
362 |   name: "fire7/relu_expand1x1"
363 |   type: "ReLU"
364 |   bottom: "fire7/expand1x1"
365 |   top: "fire7/expand1x1"
366 | }
367 | layer {
368 |   name: "fire7/expand3x3"
369 |   type: "Convolution"
370 |   bottom: "fire7/squeeze1x1"
371 |   top: "fire7/expand3x3"
372 |   convolution_param {
373 |     num_output: 192
374 |     pad: 1
375 |     kernel_size: 3
376 |   }
377 | }
378 | layer {
379 |   name: "fire7/relu_expand3x3"
380 |   type: "ReLU"
381 |   bottom: "fire7/expand3x3"
382 |   top: "fire7/expand3x3"
383 | }
384 | layer {
385 |   name: "fire7/concat"
386 |   type: "Concat"
387 |   bottom: "fire7/expand1x1"
388 |   bottom: "fire7/expand3x3"
389 |   top: "fire7/concat"
390 | }
391 | layer {
392 |   name: "fire8/squeeze1x1"
393 |   type: "Convolution"
394 |   bottom: "fire7/concat"
395 |   top: "fire8/squeeze1x1"
396 |   convolution_param {
397 |     num_output: 64
398 |     kernel_size: 1
399 |   }
400 | }
401 | layer {
402 |   name: "fire8/relu_squeeze1x1"
403 |   type: "ReLU"
404 |   bottom: "fire8/squeeze1x1"
405 |   top: "fire8/squeeze1x1"
406 | }
407 | layer {
408 |   name: "fire8/expand1x1"
409 |   type: "Convolution"
410 |   bottom: "fire8/squeeze1x1"
411 |   top: "fire8/expand1x1"
412 |   convolution_param {
413 |     num_output: 256
414 |     kernel_size: 1
415 |   }
416 | }
417 | layer {
418 |   name: "fire8/relu_expand1x1"
419 |   type: "ReLU"
420 |   bottom: "fire8/expand1x1"
421 |   top: "fire8/expand1x1"
422 | }
423 | layer {
424 |   name: "fire8/expand3x3"
425 |   type: "Convolution"
426 |   bottom: "fire8/squeeze1x1"
427 |   top: "fire8/expand3x3"
428 |   convolution_param {
429 |     num_output: 256
430 |     pad: 1
431 |     kernel_size: 3
432 |   }
433 | }
434 | layer {
435 |   name: "fire8/relu_expand3x3"
436 |   type: "ReLU"
437 |   bottom: "fire8/expand3x3"
438 |   top: "fire8/expand3x3"
439 | }
440 | layer {
441 |   name: "fire8/concat"
442 |   type: "Concat"
443 |   bottom: "fire8/expand1x1"
444 |   bottom: "fire8/expand3x3"
445 |   top: "fire8/concat"
446 | }
447 | layer {
448 |   name: "pool8"
449 |   type: "Pooling"
450 |   bottom: "fire8/concat"
451 |   top: "pool8"
452 |   pooling_param {
453 |     pool: MAX
454 |     kernel_size: 3
455 |     stride: 2
456 |  }
457 | }
458 | layer {
459 |   name: "fire9/squeeze1x1"
460 |   type: "Convolution"
461 |   bottom: "pool8"
462 |   top: "fire9/squeeze1x1"
463 |   convolution_param {
464 |     num_output: 64
465 |     kernel_size: 1
466 |   }
467 | }
468 | layer {
469 |   name: "fire9/relu_squeeze1x1"
470 |   type: "ReLU"
471 |   bottom: "fire9/squeeze1x1"
472 |   top: "fire9/squeeze1x1"
473 | }
474 | layer {
475 |   name: "fire9/expand1x1"
476 |   type: "Convolution"
477 |   bottom: "fire9/squeeze1x1"
478 |   top: "fire9/expand1x1"
479 |   convolution_param {
480 |     num_output: 256
481 |     kernel_size: 1
482 |   }
483 | }
484 | layer {
485 |   name: "fire9/relu_expand1x1"
486 |   type: "ReLU"
487 |   bottom: "fire9/expand1x1"
488 |   top: "fire9/expand1x1"
489 | }
490 | layer {
491 |   name: "fire9/expand3x3"
492 |   type: "Convolution"
493 |   bottom: "fire9/squeeze1x1"
494 |   top: "fire9/expand3x3"
495 |   convolution_param {
496 |     num_output: 256
497 |     pad: 1
498 |     kernel_size: 3
499 |   }
500 | }
501 | layer {
502 |   name: "fire9/relu_expand3x3"
503 |   type: "ReLU"
504 |   bottom: "fire9/expand3x3"
505 |   top: "fire9/expand3x3"
506 | }
507 | layer {
508 |   name: "fire9/concat"
509 |   type: "Concat"
510 |   bottom: "fire9/expand1x1"
511 |   bottom: "fire9/expand3x3"
512 |   top: "fire9/concat"
513 | }
514 | layer {
515 |   name: "drop9"
516 |   type: "Dropout"
517 |   bottom: "fire9/concat"
518 |   top: "fire9/concat"
519 |   dropout_param {
520 |     dropout_ratio: 0.5
521 |   }
522 | }
523 | layer {
524 |   name: "conv10"
525 |   type: "Convolution"
526 |   bottom: "fire9/concat"
527 |   top: "conv10"
528 |   convolution_param {
529 |     num_output: 1000
530 |     pad: 1
531 |     kernel_size: 1
532 |   }
533 | }
534 | layer {
535 |   name: "relu_conv10"
536 |   type: "ReLU"
537 |   bottom: "conv10"
538 |   top: "conv10"
539 | }
540 | layer {
541 |   name: "pool10"
542 |   type: "Pooling"
543 |   bottom: "conv10"
544 |   top: "pool10"
545 |   pooling_param {
546 |     pool: AVE
547 |     global_pooling: true
548 |   }
549 | }
550 | layer {
551 |   name: "prob"
552 |   type: "Softmax"
553 |   bottom: "pool10"
554 |   top: "prob"
555 | }
556 | 


--------------------------------------------------------------------------------
/SqueezeNet_v1.0/solver.prototxt:
--------------------------------------------------------------------------------
 1 | # please cite:
 2 | # @article{SqueezeNet,
 3 | #     Author = {Forrest N. Iandola and Matthew W. Moskewicz and Khalid Ashraf and Song Han and William J. Dally and Kurt Keutzer},
 4 | #     Title = {SqueezeNet: AlexNet-level accuracy with 50x fewer parameters and $<$1MB model size},
 5 | #     Journal = {arXiv:1602.07360},
 6 | #     Year = {2016}
 7 | # }
 8 | 
 9 | test_iter: 2000 #not subject to iter_size
10 | test_interval: 1000
11 | base_lr: 0.04
12 | display: 40
13 | max_iter: 170000
14 | iter_size: 16 #global batch size = batch_size * iter_size
15 | lr_policy: "poly"
16 | power: 1.0 #linearly decrease LR
17 | momentum: 0.9
18 | weight_decay: 0.0002
19 | snapshot: 1000
20 | snapshot_prefix: "train"
21 | solver_mode: GPU
22 | random_seed: 42
23 | net: "train_val.prototxt" #we typically do `cd SqueezeNet_v1.0; caffe train <args>` 
24 | test_initialization: false
25 | average_loss: 40
26 | 


--------------------------------------------------------------------------------
/SqueezeNet_v1.0/squeezenet_v1.0.caffemodel:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/forresti/SqueezeNet/51147dc0681638c28aad593b9e923f3500117125/SqueezeNet_v1.0/squeezenet_v1.0.caffemodel


--------------------------------------------------------------------------------
/SqueezeNet_v1.0/train_val.prototxt:
--------------------------------------------------------------------------------
  1 | # please cite:
  2 | # @article{SqueezeNet,
  3 | #     Author = {Forrest N. Iandola and Matthew W. Moskewicz and Khalid Ashraf and Song Han and William J. Dally and Kurt Keutzer},
  4 | #     Title = {SqueezeNet: AlexNet-level accuracy with 50x fewer parameters and $<$1MB model size},
  5 | #     Journal = {arXiv:1602.07360},
  6 | #     Year = {2016}
  7 | # }
  8 | layer {
  9 |   name: "data"
 10 |   type: "Data"
 11 |   top: "data"
 12 |   top: "label"
 13 |   include {
 14 |     phase: TRAIN
 15 |   }
 16 |   transform_param {
 17 |     crop_size: 227
 18 |     mean_value: 104
 19 |     mean_value: 117
 20 |     mean_value: 123
 21 |   }
 22 |   data_param {
 23 |     source: "examples/imagenet/ilsvrc12_train_lmdb"
 24 |     batch_size: 32 #*iter_size
 25 |     backend: LMDB
 26 |   }
 27 | }
 28 | layer {
 29 |   name: "data"
 30 |   type: "Data"
 31 |   top: "data"
 32 |   top: "label"
 33 |   include {
 34 |     phase: TEST
 35 |   }
 36 |   transform_param {
 37 |     crop_size: 227
 38 |     mean_value: 104
 39 |     mean_value: 117
 40 |     mean_value: 123
 41 |   }
 42 |   data_param {
 43 |     source: "examples/imagenet/ilsvrc12_val_lmdb"
 44 |     batch_size: 25 #not *iter_size
 45 |     backend: LMDB
 46 |   }
 47 | }
 48 | layer {
 49 |   name: "conv1"
 50 |   type: "Convolution"
 51 |   bottom: "data"
 52 |   top: "conv1"
 53 |   convolution_param {
 54 |     num_output: 96
 55 |     kernel_size: 7
 56 |     stride: 2
 57 |     weight_filler {
 58 |       type: "xavier"
 59 |     }
 60 |   }
 61 | }
 62 | layer {
 63 |   name: "relu_conv1"
 64 |   type: "ReLU"
 65 |   bottom: "conv1"
 66 |   top: "conv1"
 67 | }
 68 | layer {
 69 |   name: "pool1"
 70 |   type: "Pooling"
 71 |   bottom: "conv1"
 72 |   top: "pool1"
 73 |   pooling_param {
 74 |     pool: MAX
 75 |     kernel_size: 3
 76 |     stride: 2
 77 |   }
 78 | }
 79 | layer {
 80 |   name: "fire2/squeeze1x1"
 81 |   type: "Convolution"
 82 |   bottom: "pool1"
 83 |   top: "fire2/squeeze1x1"
 84 |   convolution_param {
 85 |     num_output: 16
 86 |     kernel_size: 1
 87 |     weight_filler {
 88 |       type: "xavier"
 89 |     }
 90 |   }
 91 | }
 92 | layer {
 93 |   name: "fire2/relu_squeeze1x1"
 94 |   type: "ReLU"
 95 |   bottom: "fire2/squeeze1x1"
 96 |   top: "fire2/squeeze1x1"
 97 | }
 98 | layer {
 99 |   name: "fire2/expand1x1"
100 |   type: "Convolution"
101 |   bottom: "fire2/squeeze1x1"
102 |   top: "fire2/expand1x1"
103 |   convolution_param {
104 |     num_output: 64
105 |     kernel_size: 1
106 |     weight_filler {
107 |       type: "xavier"
108 |     }
109 |   }
110 | }
111 | layer {
112 |   name: "fire2/relu_expand1x1"
113 |   type: "ReLU"
114 |   bottom: "fire2/expand1x1"
115 |   top: "fire2/expand1x1"
116 | }
117 | layer {
118 |   name: "fire2/expand3x3"
119 |   type: "Convolution"
120 |   bottom: "fire2/squeeze1x1"
121 |   top: "fire2/expand3x3"
122 |   convolution_param {
123 |     num_output: 64
124 |     pad: 1
125 |     kernel_size: 3
126 |     weight_filler {
127 |       type: "xavier"
128 |     }
129 |   }
130 | }
131 | layer {
132 |   name: "fire2/relu_expand3x3"
133 |   type: "ReLU"
134 |   bottom: "fire2/expand3x3"
135 |   top: "fire2/expand3x3"
136 | }
137 | layer {
138 |   name: "fire2/concat"
139 |   type: "Concat"
140 |   bottom: "fire2/expand1x1"
141 |   bottom: "fire2/expand3x3"
142 |   top: "fire2/concat"
143 | }
144 | layer {
145 |   name: "fire3/squeeze1x1"
146 |   type: "Convolution"
147 |   bottom: "fire2/concat"
148 |   top: "fire3/squeeze1x1"
149 |   convolution_param {
150 |     num_output: 16
151 |     kernel_size: 1
152 |     weight_filler {
153 |       type: "xavier"
154 |     }
155 |   }
156 | }
157 | layer {
158 |   name: "fire3/relu_squeeze1x1"
159 |   type: "ReLU"
160 |   bottom: "fire3/squeeze1x1"
161 |   top: "fire3/squeeze1x1"
162 | }
163 | layer {
164 |   name: "fire3/expand1x1"
165 |   type: "Convolution"
166 |   bottom: "fire3/squeeze1x1"
167 |   top: "fire3/expand1x1"
168 |   convolution_param {
169 |     num_output: 64
170 |     kernel_size: 1
171 |     weight_filler {
172 |       type: "xavier"
173 |     }
174 |   }
175 | }
176 | layer {
177 |   name: "fire3/relu_expand1x1"
178 |   type: "ReLU"
179 |   bottom: "fire3/expand1x1"
180 |   top: "fire3/expand1x1"
181 | }
182 | layer {
183 |   name: "fire3/expand3x3"
184 |   type: "Convolution"
185 |   bottom: "fire3/squeeze1x1"
186 |   top: "fire3/expand3x3"
187 |   convolution_param {
188 |     num_output: 64
189 |     pad: 1
190 |     kernel_size: 3
191 |     weight_filler {
192 |       type: "xavier"
193 |     }
194 |   }
195 | }
196 | layer {
197 |   name: "fire3/relu_expand3x3"
198 |   type: "ReLU"
199 |   bottom: "fire3/expand3x3"
200 |   top: "fire3/expand3x3"
201 | }
202 | layer {
203 |   name: "fire3/concat"
204 |   type: "Concat"
205 |   bottom: "fire3/expand1x1"
206 |   bottom: "fire3/expand3x3"
207 |   top: "fire3/concat"
208 | }
209 | layer {
210 |   name: "fire4/squeeze1x1"
211 |   type: "Convolution"
212 |   bottom: "fire3/concat"
213 |   top: "fire4/squeeze1x1"
214 |   convolution_param {
215 |     num_output: 32
216 |     kernel_size: 1
217 |     weight_filler {
218 |       type: "xavier"
219 |     }
220 |   }
221 | }
222 | layer {
223 |   name: "fire4/relu_squeeze1x1"
224 |   type: "ReLU"
225 |   bottom: "fire4/squeeze1x1"
226 |   top: "fire4/squeeze1x1"
227 | }
228 | layer {
229 |   name: "fire4/expand1x1"
230 |   type: "Convolution"
231 |   bottom: "fire4/squeeze1x1"
232 |   top: "fire4/expand1x1"
233 |   convolution_param {
234 |     num_output: 128
235 |     kernel_size: 1
236 |     weight_filler {
237 |       type: "xavier"
238 |     }
239 |   }
240 | }
241 | layer {
242 |   name: "fire4/relu_expand1x1"
243 |   type: "ReLU"
244 |   bottom: "fire4/expand1x1"
245 |   top: "fire4/expand1x1"
246 | }
247 | layer {
248 |   name: "fire4/expand3x3"
249 |   type: "Convolution"
250 |   bottom: "fire4/squeeze1x1"
251 |   top: "fire4/expand3x3"
252 |   convolution_param {
253 |     num_output: 128
254 |     pad: 1
255 |     kernel_size: 3
256 |     weight_filler {
257 |       type: "xavier"
258 |     }
259 |   }
260 | }
261 | layer {
262 |   name: "fire4/relu_expand3x3"
263 |   type: "ReLU"
264 |   bottom: "fire4/expand3x3"
265 |   top: "fire4/expand3x3"
266 | }
267 | layer {
268 |   name: "fire4/concat"
269 |   type: "Concat"
270 |   bottom: "fire4/expand1x1"
271 |   bottom: "fire4/expand3x3"
272 |   top: "fire4/concat"
273 | }
274 | layer {
275 |   name: "pool4"
276 |   type: "Pooling"
277 |   bottom: "fire4/concat"
278 |   top: "pool4"
279 |   pooling_param {
280 |     pool: MAX
281 |     kernel_size: 3
282 |     stride: 2
283 |   }
284 | }
285 | layer {
286 |   name: "fire5/squeeze1x1"
287 |   type: "Convolution"
288 |   bottom: "pool4"
289 |   top: "fire5/squeeze1x1"
290 |   convolution_param {
291 |     num_output: 32
292 |     kernel_size: 1
293 |     weight_filler {
294 |       type: "xavier"
295 |     }
296 |   }
297 | }
298 | layer {
299 |   name: "fire5/relu_squeeze1x1"
300 |   type: "ReLU"
301 |   bottom: "fire5/squeeze1x1"
302 |   top: "fire5/squeeze1x1"
303 | }
304 | layer {
305 |   name: "fire5/expand1x1"
306 |   type: "Convolution"
307 |   bottom: "fire5/squeeze1x1"
308 |   top: "fire5/expand1x1"
309 |   convolution_param {
310 |     num_output: 128
311 |     kernel_size: 1
312 |     weight_filler {
313 |       type: "xavier"
314 |     }
315 |   }
316 | }
317 | layer {
318 |   name: "fire5/relu_expand1x1"
319 |   type: "ReLU"
320 |   bottom: "fire5/expand1x1"
321 |   top: "fire5/expand1x1"
322 | }
323 | layer {
324 |   name: "fire5/expand3x3"
325 |   type: "Convolution"
326 |   bottom: "fire5/squeeze1x1"
327 |   top: "fire5/expand3x3"
328 |   convolution_param {
329 |     num_output: 128
330 |     pad: 1
331 |     kernel_size: 3
332 |     weight_filler {
333 |       type: "xavier"
334 |     }
335 |   }
336 | }
337 | layer {
338 |   name: "fire5/relu_expand3x3"
339 |   type: "ReLU"
340 |   bottom: "fire5/expand3x3"
341 |   top: "fire5/expand3x3"
342 | }
343 | layer {
344 |   name: "fire5/concat"
345 |   type: "Concat"
346 |   bottom: "fire5/expand1x1"
347 |   bottom: "fire5/expand3x3"
348 |   top: "fire5/concat"
349 | }
350 | layer {
351 |   name: "fire6/squeeze1x1"
352 |   type: "Convolution"
353 |   bottom: "fire5/concat"
354 |   top: "fire6/squeeze1x1"
355 |   convolution_param {
356 |     num_output: 48
357 |     kernel_size: 1
358 |     weight_filler {
359 |       type: "xavier"
360 |     }
361 |   }
362 | }
363 | layer {
364 |   name: "fire6/relu_squeeze1x1"
365 |   type: "ReLU"
366 |   bottom: "fire6/squeeze1x1"
367 |   top: "fire6/squeeze1x1"
368 | }
369 | layer {
370 |   name: "fire6/expand1x1"
371 |   type: "Convolution"
372 |   bottom: "fire6/squeeze1x1"
373 |   top: "fire6/expand1x1"
374 |   convolution_param {
375 |     num_output: 192
376 |     kernel_size: 1
377 |     weight_filler {
378 |       type: "xavier"
379 |     }
380 |   }
381 | }
382 | layer {
383 |   name: "fire6/relu_expand1x1"
384 |   type: "ReLU"
385 |   bottom: "fire6/expand1x1"
386 |   top: "fire6/expand1x1"
387 | }
388 | layer {
389 |   name: "fire6/expand3x3"
390 |   type: "Convolution"
391 |   bottom: "fire6/squeeze1x1"
392 |   top: "fire6/expand3x3"
393 |   convolution_param {
394 |     num_output: 192
395 |     pad: 1
396 |     kernel_size: 3
397 |     weight_filler {
398 |       type: "xavier"
399 |     }
400 |   }
401 | }
402 | layer {
403 |   name: "fire6/relu_expand3x3"
404 |   type: "ReLU"
405 |   bottom: "fire6/expand3x3"
406 |   top: "fire6/expand3x3"
407 | }
408 | layer {
409 |   name: "fire6/concat"
410 |   type: "Concat"
411 |   bottom: "fire6/expand1x1"
412 |   bottom: "fire6/expand3x3"
413 |   top: "fire6/concat"
414 | }
415 | layer {
416 |   name: "fire7/squeeze1x1"
417 |   type: "Convolution"
418 |   bottom: "fire6/concat"
419 |   top: "fire7/squeeze1x1"
420 |   convolution_param {
421 |     num_output: 48
422 |     kernel_size: 1
423 |     weight_filler {
424 |       type: "xavier"
425 |     }
426 |   }
427 | }
428 | layer {
429 |   name: "fire7/relu_squeeze1x1"
430 |   type: "ReLU"
431 |   bottom: "fire7/squeeze1x1"
432 |   top: "fire7/squeeze1x1"
433 | }
434 | layer {
435 |   name: "fire7/expand1x1"
436 |   type: "Convolution"
437 |   bottom: "fire7/squeeze1x1"
438 |   top: "fire7/expand1x1"
439 |   convolution_param {
440 |     num_output: 192
441 |     kernel_size: 1
442 |     weight_filler {
443 |       type: "xavier"
444 |     }
445 |   }
446 | }
447 | layer {
448 |   name: "fire7/relu_expand1x1"
449 |   type: "ReLU"
450 |   bottom: "fire7/expand1x1"
451 |   top: "fire7/expand1x1"
452 | }
453 | layer {
454 |   name: "fire7/expand3x3"
455 |   type: "Convolution"
456 |   bottom: "fire7/squeeze1x1"
457 |   top: "fire7/expand3x3"
458 |   convolution_param {
459 |     num_output: 192
460 |     pad: 1
461 |     kernel_size: 3
462 |     weight_filler {
463 |       type: "xavier"
464 |     }
465 |   }
466 | }
467 | layer {
468 |   name: "fire7/relu_expand3x3"
469 |   type: "ReLU"
470 |   bottom: "fire7/expand3x3"
471 |   top: "fire7/expand3x3"
472 | }
473 | layer {
474 |   name: "fire7/concat"
475 |   type: "Concat"
476 |   bottom: "fire7/expand1x1"
477 |   bottom: "fire7/expand3x3"
478 |   top: "fire7/concat"
479 | }
480 | layer {
481 |   name: "fire8/squeeze1x1"
482 |   type: "Convolution"
483 |   bottom: "fire7/concat"
484 |   top: "fire8/squeeze1x1"
485 |   convolution_param {
486 |     num_output: 64
487 |     kernel_size: 1
488 |     weight_filler {
489 |       type: "xavier"
490 |     }
491 |   }
492 | }
493 | layer {
494 |   name: "fire8/relu_squeeze1x1"
495 |   type: "ReLU"
496 |   bottom: "fire8/squeeze1x1"
497 |   top: "fire8/squeeze1x1"
498 | }
499 | layer {
500 |   name: "fire8/expand1x1"
501 |   type: "Convolution"
502 |   bottom: "fire8/squeeze1x1"
503 |   top: "fire8/expand1x1"
504 |   convolution_param {
505 |     num_output: 256
506 |     kernel_size: 1
507 |     weight_filler {
508 |       type: "xavier"
509 |     }
510 |   }
511 | }
512 | layer {
513 |   name: "fire8/relu_expand1x1"
514 |   type: "ReLU"
515 |   bottom: "fire8/expand1x1"
516 |   top: "fire8/expand1x1"
517 | }
518 | layer {
519 |   name: "fire8/expand3x3"
520 |   type: "Convolution"
521 |   bottom: "fire8/squeeze1x1"
522 |   top: "fire8/expand3x3"
523 |   convolution_param {
524 |     num_output: 256
525 |     pad: 1
526 |     kernel_size: 3
527 |     weight_filler {
528 |       type: "xavier"
529 |     }
530 |   }
531 | }
532 | layer {
533 |   name: "fire8/relu_expand3x3"
534 |   type: "ReLU"
535 |   bottom: "fire8/expand3x3"
536 |   top: "fire8/expand3x3"
537 | }
538 | layer {
539 |   name: "fire8/concat"
540 |   type: "Concat"
541 |   bottom: "fire8/expand1x1"
542 |   bottom: "fire8/expand3x3"
543 |   top: "fire8/concat"
544 | }
545 | layer {
546 |   name: "pool8"
547 |   type: "Pooling"
548 |   bottom: "fire8/concat"
549 |   top: "pool8"
550 |   pooling_param {
551 |     pool: MAX
552 |     kernel_size: 3
553 |     stride: 2
554 |   }
555 | }
556 | layer {
557 |   name: "fire9/squeeze1x1"
558 |   type: "Convolution"
559 |   bottom: "pool8"
560 |   top: "fire9/squeeze1x1"
561 |   convolution_param {
562 |     num_output: 64
563 |     kernel_size: 1
564 |     weight_filler {
565 |       type: "xavier"
566 |     }
567 |   }
568 | }
569 | layer {
570 |   name: "fire9/relu_squeeze1x1"
571 |   type: "ReLU"
572 |   bottom: "fire9/squeeze1x1"
573 |   top: "fire9/squeeze1x1"
574 | }
575 | layer {
576 |   name: "fire9/expand1x1"
577 |   type: "Convolution"
578 |   bottom: "fire9/squeeze1x1"
579 |   top: "fire9/expand1x1"
580 |   convolution_param {
581 |     num_output: 256
582 |     kernel_size: 1
583 |     weight_filler {
584 |       type: "xavier"
585 |     }
586 |   }
587 | }
588 | layer {
589 |   name: "fire9/relu_expand1x1"
590 |   type: "ReLU"
591 |   bottom: "fire9/expand1x1"
592 |   top: "fire9/expand1x1"
593 | }
594 | layer {
595 |   name: "fire9/expand3x3"
596 |   type: "Convolution"
597 |   bottom: "fire9/squeeze1x1"
598 |   top: "fire9/expand3x3"
599 |   convolution_param {
600 |     num_output: 256
601 |     pad: 1
602 |     kernel_size: 3
603 |     weight_filler {
604 |       type: "xavier"
605 |     }
606 |   }
607 | }
608 | layer {
609 |   name: "fire9/relu_expand3x3"
610 |   type: "ReLU"
611 |   bottom: "fire9/expand3x3"
612 |   top: "fire9/expand3x3"
613 | }
614 | layer {
615 |   name: "fire9/concat"
616 |   type: "Concat"
617 |   bottom: "fire9/expand1x1"
618 |   bottom: "fire9/expand3x3"
619 |   top: "fire9/concat"
620 | }
621 | layer {
622 |   name: "drop9"
623 |   type: "Dropout"
624 |   bottom: "fire9/concat"
625 |   top: "fire9/concat"
626 |   dropout_param {
627 |     dropout_ratio: 0.5
628 |   }
629 | }
630 | layer {
631 |   name: "conv10"
632 |   type: "Convolution"
633 |   bottom: "fire9/concat"
634 |   top: "conv10"
635 |   convolution_param {
636 |     num_output: 1000
637 |     pad: 1
638 |     kernel_size: 1
639 |     weight_filler {
640 |       type: "gaussian"
641 |       mean: 0.0
642 |       std: 0.01
643 |     }
644 |   }
645 | }
646 | layer {
647 |   name: "relu_conv10"
648 |   type: "ReLU"
649 |   bottom: "conv10"
650 |   top: "conv10"
651 | }
652 | layer {
653 |   name: "pool10"
654 |   type: "Pooling"
655 |   bottom: "conv10"
656 |   top: "pool10"
657 |   pooling_param {
658 |     pool: AVE
659 |     global_pooling: true
660 |   }
661 | }
662 | layer {
663 |   name: "loss"
664 |   type: "SoftmaxWithLoss"
665 |   bottom: "pool10"
666 |   bottom: "label"
667 |   top: "loss"
668 |   include {
669 |     phase: TRAIN
670 |   }
671 | }
672 | layer {
673 |   name: "accuracy"
674 |   type: "Accuracy"
675 |   bottom: "pool10"
676 |   bottom: "label"
677 |   top: "accuracy"
678 |   include {
679 |     phase: TEST
680 |   }
681 | }
682 | layer {
683 |   name: "accuracy_top5"
684 |   type: "Accuracy"
685 |   bottom: "pool10"
686 |   bottom: "label"
687 |   top: "accuracy_top5"
688 |   include {
689 |     phase: TEST
690 |   }
691 |   accuracy_param {
692 |     top_k: 5
693 |   }
694 | }
695 | 


--------------------------------------------------------------------------------
/SqueezeNet_v1.1/README.md:
--------------------------------------------------------------------------------
 1 | 
 2 | **What's new in SqueezeNet v1.1?**
 3 | 
 4 | |                 | SqueezeNet v1.0                  | SqueezeNet v1.1                  |
 5 | | :------------- |:-------------:| :-----:|
 6 | | conv1:          | 96 filters of resolution 7x7     | 64 filters of resolution 3x3     |
 7 | | pooling layers: | pool_{1,4,8}                     | pool_{1,3,5}                     |
 8 | | computation     | 1.72 GFLOPS/image                | 0.72 GFLOPS/image: *2.4x less computation* |
 9 | | ImageNet accuracy        | >= 80.3% top-5                   | >= 80.3% top-5                   |    
10 | 
11 | 
12 | SqueezeNet v1.1 has 2.4x less computation than v1.0, without sacrificing accuracy. 
13 | 


--------------------------------------------------------------------------------
/SqueezeNet_v1.1/deploy.prototxt:
--------------------------------------------------------------------------------
  1 | # please cite:
  2 | # @article{SqueezeNet,
  3 | #     Author = {Forrest N. Iandola and Matthew W. Moskewicz and Khalid Ashraf and Song Han and William J. Dally and Kurt Keutzer},
  4 | #     Title = {SqueezeNet: AlexNet-level accuracy with 50x fewer parameters and $<$1MB model size},
  5 | #     Journal = {arXiv:1602.07360},
  6 | #     Year = {2016}
  7 | # }
  8 | 
  9 | layer {
 10 |   name: "data"
 11 |   type: "Input"
 12 |   top: "data"
 13 |   input_param { shape: { dim: 10 dim: 3 dim: 227 dim: 227 } }
 14 | }
 15 | layer {
 16 |   name: "conv1"
 17 |   type: "Convolution"
 18 |   bottom: "data"
 19 |   top: "conv1"
 20 |   convolution_param {
 21 |     num_output: 64
 22 |     kernel_size: 3
 23 |     stride: 2
 24 |   }
 25 | }
 26 | layer {
 27 |   name: "relu_conv1"
 28 |   type: "ReLU"
 29 |   bottom: "conv1"
 30 |   top: "conv1"
 31 | }
 32 | layer {
 33 |   name: "pool1"
 34 |   type: "Pooling"
 35 |   bottom: "conv1"
 36 |   top: "pool1"
 37 |   pooling_param {
 38 |     pool: MAX
 39 |     kernel_size: 3
 40 |     stride: 2
 41 |   }
 42 | }
 43 | layer {
 44 |   name: "fire2/squeeze1x1"
 45 |   type: "Convolution"
 46 |   bottom: "pool1"
 47 |   top: "fire2/squeeze1x1"
 48 |   convolution_param {
 49 |     num_output: 16
 50 |     kernel_size: 1
 51 |   }
 52 | }
 53 | layer {
 54 |   name: "fire2/relu_squeeze1x1"
 55 |   type: "ReLU"
 56 |   bottom: "fire2/squeeze1x1"
 57 |   top: "fire2/squeeze1x1"
 58 | }
 59 | layer {
 60 |   name: "fire2/expand1x1"
 61 |   type: "Convolution"
 62 |   bottom: "fire2/squeeze1x1"
 63 |   top: "fire2/expand1x1"
 64 |   convolution_param {
 65 |     num_output: 64
 66 |     kernel_size: 1
 67 |   }
 68 | }
 69 | layer {
 70 |   name: "fire2/relu_expand1x1"
 71 |   type: "ReLU"
 72 |   bottom: "fire2/expand1x1"
 73 |   top: "fire2/expand1x1"
 74 | }
 75 | layer {
 76 |   name: "fire2/expand3x3"
 77 |   type: "Convolution"
 78 |   bottom: "fire2/squeeze1x1"
 79 |   top: "fire2/expand3x3"
 80 |   convolution_param {
 81 |     num_output: 64
 82 |     pad: 1
 83 |     kernel_size: 3
 84 |   }
 85 | }
 86 | layer {
 87 |   name: "fire2/relu_expand3x3"
 88 |   type: "ReLU"
 89 |   bottom: "fire2/expand3x3"
 90 |   top: "fire2/expand3x3"
 91 | }
 92 | layer {
 93 |   name: "fire2/concat"
 94 |   type: "Concat"
 95 |   bottom: "fire2/expand1x1"
 96 |   bottom: "fire2/expand3x3"
 97 |   top: "fire2/concat"
 98 | }
 99 | layer {
100 |   name: "fire3/squeeze1x1"
101 |   type: "Convolution"
102 |   bottom: "fire2/concat"
103 |   top: "fire3/squeeze1x1"
104 |   convolution_param {
105 |     num_output: 16
106 |     kernel_size: 1
107 |   }
108 | }
109 | layer {
110 |   name: "fire3/relu_squeeze1x1"
111 |   type: "ReLU"
112 |   bottom: "fire3/squeeze1x1"
113 |   top: "fire3/squeeze1x1"
114 | }
115 | layer {
116 |   name: "fire3/expand1x1"
117 |   type: "Convolution"
118 |   bottom: "fire3/squeeze1x1"
119 |   top: "fire3/expand1x1"
120 |   convolution_param {
121 |     num_output: 64
122 |     kernel_size: 1
123 |   }
124 | }
125 | layer {
126 |   name: "fire3/relu_expand1x1"
127 |   type: "ReLU"
128 |   bottom: "fire3/expand1x1"
129 |   top: "fire3/expand1x1"
130 | }
131 | layer {
132 |   name: "fire3/expand3x3"
133 |   type: "Convolution"
134 |   bottom: "fire3/squeeze1x1"
135 |   top: "fire3/expand3x3"
136 |   convolution_param {
137 |     num_output: 64
138 |     pad: 1
139 |     kernel_size: 3
140 |   }
141 | }
142 | layer {
143 |   name: "fire3/relu_expand3x3"
144 |   type: "ReLU"
145 |   bottom: "fire3/expand3x3"
146 |   top: "fire3/expand3x3"
147 | }
148 | layer {
149 |   name: "fire3/concat"
150 |   type: "Concat"
151 |   bottom: "fire3/expand1x1"
152 |   bottom: "fire3/expand3x3"
153 |   top: "fire3/concat"
154 | }
155 | layer {
156 |   name: "pool3"
157 |   type: "Pooling"
158 |   bottom: "fire3/concat"
159 |   top: "pool3"
160 |   pooling_param {
161 |     pool: MAX
162 |     kernel_size: 3
163 |     stride: 2
164 |   }
165 | }
166 | layer {
167 |   name: "fire4/squeeze1x1"
168 |   type: "Convolution"
169 |   bottom: "pool3"
170 |   top: "fire4/squeeze1x1"
171 |   convolution_param {
172 |     num_output: 32
173 |     kernel_size: 1
174 |   }
175 | }
176 | layer {
177 |   name: "fire4/relu_squeeze1x1"
178 |   type: "ReLU"
179 |   bottom: "fire4/squeeze1x1"
180 |   top: "fire4/squeeze1x1"
181 | }
182 | layer {
183 |   name: "fire4/expand1x1"
184 |   type: "Convolution"
185 |   bottom: "fire4/squeeze1x1"
186 |   top: "fire4/expand1x1"
187 |   convolution_param {
188 |     num_output: 128
189 |     kernel_size: 1
190 |   }
191 | }
192 | layer {
193 |   name: "fire4/relu_expand1x1"
194 |   type: "ReLU"
195 |   bottom: "fire4/expand1x1"
196 |   top: "fire4/expand1x1"
197 | }
198 | layer {
199 |   name: "fire4/expand3x3"
200 |   type: "Convolution"
201 |   bottom: "fire4/squeeze1x1"
202 |   top: "fire4/expand3x3"
203 |   convolution_param {
204 |     num_output: 128
205 |     pad: 1
206 |     kernel_size: 3
207 |   }
208 | }
209 | layer {
210 |   name: "fire4/relu_expand3x3"
211 |   type: "ReLU"
212 |   bottom: "fire4/expand3x3"
213 |   top: "fire4/expand3x3"
214 | }
215 | layer {
216 |   name: "fire4/concat"
217 |   type: "Concat"
218 |   bottom: "fire4/expand1x1"
219 |   bottom: "fire4/expand3x3"
220 |   top: "fire4/concat"
221 | }
222 | layer {
223 |   name: "fire5/squeeze1x1"
224 |   type: "Convolution"
225 |   bottom: "fire4/concat"
226 |   top: "fire5/squeeze1x1"
227 |   convolution_param {
228 |     num_output: 32
229 |     kernel_size: 1
230 |   }
231 | }
232 | layer {
233 |   name: "fire5/relu_squeeze1x1"
234 |   type: "ReLU"
235 |   bottom: "fire5/squeeze1x1"
236 |   top: "fire5/squeeze1x1"
237 | }
238 | layer {
239 |   name: "fire5/expand1x1"
240 |   type: "Convolution"
241 |   bottom: "fire5/squeeze1x1"
242 |   top: "fire5/expand1x1"
243 |   convolution_param {
244 |     num_output: 128
245 |     kernel_size: 1
246 |   }
247 | }
248 | layer {
249 |   name: "fire5/relu_expand1x1"
250 |   type: "ReLU"
251 |   bottom: "fire5/expand1x1"
252 |   top: "fire5/expand1x1"
253 | }
254 | layer {
255 |   name: "fire5/expand3x3"
256 |   type: "Convolution"
257 |   bottom: "fire5/squeeze1x1"
258 |   top: "fire5/expand3x3"
259 |   convolution_param {
260 |     num_output: 128
261 |     pad: 1
262 |     kernel_size: 3
263 |   }
264 | }
265 | layer {
266 |   name: "fire5/relu_expand3x3"
267 |   type: "ReLU"
268 |   bottom: "fire5/expand3x3"
269 |   top: "fire5/expand3x3"
270 | }
271 | layer {
272 |   name: "fire5/concat"
273 |   type: "Concat"
274 |   bottom: "fire5/expand1x1"
275 |   bottom: "fire5/expand3x3"
276 |   top: "fire5/concat"
277 | }
278 | layer {
279 |   name: "pool5"
280 |   type: "Pooling"
281 |   bottom: "fire5/concat"
282 |   top: "pool5"
283 |   pooling_param {
284 |     pool: MAX
285 |     kernel_size: 3
286 |     stride: 2
287 |   }
288 | }
289 | layer {
290 |   name: "fire6/squeeze1x1"
291 |   type: "Convolution"
292 |   bottom: "pool5"
293 |   top: "fire6/squeeze1x1"
294 |   convolution_param {
295 |     num_output: 48
296 |     kernel_size: 1
297 |   }
298 | }
299 | layer {
300 |   name: "fire6/relu_squeeze1x1"
301 |   type: "ReLU"
302 |   bottom: "fire6/squeeze1x1"
303 |   top: "fire6/squeeze1x1"
304 | }
305 | layer {
306 |   name: "fire6/expand1x1"
307 |   type: "Convolution"
308 |   bottom: "fire6/squeeze1x1"
309 |   top: "fire6/expand1x1"
310 |   convolution_param {
311 |     num_output: 192
312 |     kernel_size: 1
313 |   }
314 | }
315 | layer {
316 |   name: "fire6/relu_expand1x1"
317 |   type: "ReLU"
318 |   bottom: "fire6/expand1x1"
319 |   top: "fire6/expand1x1"
320 | }
321 | layer {
322 |   name: "fire6/expand3x3"
323 |   type: "Convolution"
324 |   bottom: "fire6/squeeze1x1"
325 |   top: "fire6/expand3x3"
326 |   convolution_param {
327 |     num_output: 192
328 |     pad: 1
329 |     kernel_size: 3
330 |   }
331 | }
332 | layer {
333 |   name: "fire6/relu_expand3x3"
334 |   type: "ReLU"
335 |   bottom: "fire6/expand3x3"
336 |   top: "fire6/expand3x3"
337 | }
338 | layer {
339 |   name: "fire6/concat"
340 |   type: "Concat"
341 |   bottom: "fire6/expand1x1"
342 |   bottom: "fire6/expand3x3"
343 |   top: "fire6/concat"
344 | }
345 | layer {
346 |   name: "fire7/squeeze1x1"
347 |   type: "Convolution"
348 |   bottom: "fire6/concat"
349 |   top: "fire7/squeeze1x1"
350 |   convolution_param {
351 |     num_output: 48
352 |     kernel_size: 1
353 |   }
354 | }
355 | layer {
356 |   name: "fire7/relu_squeeze1x1"
357 |   type: "ReLU"
358 |   bottom: "fire7/squeeze1x1"
359 |   top: "fire7/squeeze1x1"
360 | }
361 | layer {
362 |   name: "fire7/expand1x1"
363 |   type: "Convolution"
364 |   bottom: "fire7/squeeze1x1"
365 |   top: "fire7/expand1x1"
366 |   convolution_param {
367 |     num_output: 192
368 |     kernel_size: 1
369 |   }
370 | }
371 | layer {
372 |   name: "fire7/relu_expand1x1"
373 |   type: "ReLU"
374 |   bottom: "fire7/expand1x1"
375 |   top: "fire7/expand1x1"
376 | }
377 | layer {
378 |   name: "fire7/expand3x3"
379 |   type: "Convolution"
380 |   bottom: "fire7/squeeze1x1"
381 |   top: "fire7/expand3x3"
382 |   convolution_param {
383 |     num_output: 192
384 |     pad: 1
385 |     kernel_size: 3
386 |   }
387 | }
388 | layer {
389 |   name: "fire7/relu_expand3x3"
390 |   type: "ReLU"
391 |   bottom: "fire7/expand3x3"
392 |   top: "fire7/expand3x3"
393 | }
394 | layer {
395 |   name: "fire7/concat"
396 |   type: "Concat"
397 |   bottom: "fire7/expand1x1"
398 |   bottom: "fire7/expand3x3"
399 |   top: "fire7/concat"
400 | }
401 | layer {
402 |   name: "fire8/squeeze1x1"
403 |   type: "Convolution"
404 |   bottom: "fire7/concat"
405 |   top: "fire8/squeeze1x1"
406 |   convolution_param {
407 |     num_output: 64
408 |     kernel_size: 1
409 |   }
410 | }
411 | layer {
412 |   name: "fire8/relu_squeeze1x1"
413 |   type: "ReLU"
414 |   bottom: "fire8/squeeze1x1"
415 |   top: "fire8/squeeze1x1"
416 | }
417 | layer {
418 |   name: "fire8/expand1x1"
419 |   type: "Convolution"
420 |   bottom: "fire8/squeeze1x1"
421 |   top: "fire8/expand1x1"
422 |   convolution_param {
423 |     num_output: 256
424 |     kernel_size: 1
425 |   }
426 | }
427 | layer {
428 |   name: "fire8/relu_expand1x1"
429 |   type: "ReLU"
430 |   bottom: "fire8/expand1x1"
431 |   top: "fire8/expand1x1"
432 | }
433 | layer {
434 |   name: "fire8/expand3x3"
435 |   type: "Convolution"
436 |   bottom: "fire8/squeeze1x1"
437 |   top: "fire8/expand3x3"
438 |   convolution_param {
439 |     num_output: 256
440 |     pad: 1
441 |     kernel_size: 3
442 |   }
443 | }
444 | layer {
445 |   name: "fire8/relu_expand3x3"
446 |   type: "ReLU"
447 |   bottom: "fire8/expand3x3"
448 |   top: "fire8/expand3x3"
449 | }
450 | layer {
451 |   name: "fire8/concat"
452 |   type: "Concat"
453 |   bottom: "fire8/expand1x1"
454 |   bottom: "fire8/expand3x3"
455 |   top: "fire8/concat"
456 | }
457 | layer {
458 |   name: "fire9/squeeze1x1"
459 |   type: "Convolution"
460 |   bottom: "fire8/concat"
461 |   top: "fire9/squeeze1x1"
462 |   convolution_param {
463 |     num_output: 64
464 |     kernel_size: 1
465 |   }
466 | }
467 | layer {
468 |   name: "fire9/relu_squeeze1x1"
469 |   type: "ReLU"
470 |   bottom: "fire9/squeeze1x1"
471 |   top: "fire9/squeeze1x1"
472 | }
473 | layer {
474 |   name: "fire9/expand1x1"
475 |   type: "Convolution"
476 |   bottom: "fire9/squeeze1x1"
477 |   top: "fire9/expand1x1"
478 |   convolution_param {
479 |     num_output: 256
480 |     kernel_size: 1
481 |   }
482 | }
483 | layer {
484 |   name: "fire9/relu_expand1x1"
485 |   type: "ReLU"
486 |   bottom: "fire9/expand1x1"
487 |   top: "fire9/expand1x1"
488 | }
489 | layer {
490 |   name: "fire9/expand3x3"
491 |   type: "Convolution"
492 |   bottom: "fire9/squeeze1x1"
493 |   top: "fire9/expand3x3"
494 |   convolution_param {
495 |     num_output: 256
496 |     pad: 1
497 |     kernel_size: 3
498 |   }
499 | }
500 | layer {
501 |   name: "fire9/relu_expand3x3"
502 |   type: "ReLU"
503 |   bottom: "fire9/expand3x3"
504 |   top: "fire9/expand3x3"
505 | }
506 | layer {
507 |   name: "fire9/concat"
508 |   type: "Concat"
509 |   bottom: "fire9/expand1x1"
510 |   bottom: "fire9/expand3x3"
511 |   top: "fire9/concat"
512 | }
513 | layer {
514 |   name: "drop9"
515 |   type: "Dropout"
516 |   bottom: "fire9/concat"
517 |   top: "fire9/concat"
518 |   dropout_param {
519 |     dropout_ratio: 0.5
520 |   }
521 | }
522 | layer {
523 |   name: "conv10"
524 |   type: "Convolution"
525 |   bottom: "fire9/concat"
526 |   top: "conv10"
527 |   convolution_param {
528 |     num_output: 1000
529 |     kernel_size: 1
530 |   }
531 | }
532 | layer {
533 |   name: "relu_conv10"
534 |   type: "ReLU"
535 |   bottom: "conv10"
536 |   top: "conv10"
537 | }
538 | layer {
539 |   name: "pool10"
540 |   type: "Pooling"
541 |   bottom: "conv10"
542 |   top: "pool10"
543 |   pooling_param {
544 |     pool: AVE
545 |     global_pooling: true
546 |   }
547 | }
548 | layer {
549 |   name: "prob"
550 |   type: "Softmax"
551 |   bottom: "pool10"
552 |   top: "prob"
553 | }
554 | 


--------------------------------------------------------------------------------
/SqueezeNet_v1.1/solver.prototxt:
--------------------------------------------------------------------------------
 1 | # please cite:
 2 | # @article{SqueezeNet,
 3 | #     Author = {Forrest N. Iandola and Matthew W. Moskewicz and Khalid Ashraf and Song Han and William J. Dally and Kurt Keutzer},
 4 | #     Title = {SqueezeNet: AlexNet-level accuracy with 50x fewer parameters and $<$1MB model size},
 5 | #     Journal = {arXiv:1602.07360},
 6 | #     Year = {2016}
 7 | # }
 8 | 
 9 | test_iter: 2000 #not subject to iter_size
10 | test_interval: 1000
11 | base_lr: 0.04
12 | display: 40
13 | max_iter: 170000
14 | iter_size: 16 #global batch size = batch_size * iter_size
15 | lr_policy: "poly"
16 | power: 1.0 #linearly decrease LR
17 | momentum: 0.9
18 | weight_decay: 0.0002
19 | snapshot: 1000
20 | snapshot_prefix: "train"
21 | solver_mode: GPU
22 | random_seed: 42
23 | net: "train_val.prototxt" #we typically do `cd SqueezeNet_v1.0; caffe train <args>` 
24 | test_initialization: false
25 | average_loss: 40
26 | 


--------------------------------------------------------------------------------
/SqueezeNet_v1.1/squeezenet_v1.1.caffemodel:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/forresti/SqueezeNet/51147dc0681638c28aad593b9e923f3500117125/SqueezeNet_v1.1/squeezenet_v1.1.caffemodel


--------------------------------------------------------------------------------
/SqueezeNet_v1.1/train_val.prototxt:
--------------------------------------------------------------------------------
  1 | # please cite:
  2 | # @article{SqueezeNet,
  3 | #     Author = {Forrest N. Iandola and Matthew W. Moskewicz and Khalid Ashraf and Song Han and William J. Dally and Kurt Keutzer},
  4 | #     Title = {SqueezeNet: AlexNet-level accuracy with 50x fewer parameters and $<$1MB model size},
  5 | #     Journal = {arXiv:1602.07360},
  6 | #     Year = {2016}
  7 | # }
  8 | layer {
  9 |   name: "data"
 10 |   type: "Data"
 11 |   top: "data"
 12 |   top: "label"
 13 |   include {
 14 |     phase: TRAIN
 15 |   }
 16 |   transform_param {
 17 |     crop_size: 227
 18 |     mean_value: 104
 19 |     mean_value: 117
 20 |     mean_value: 123
 21 |   }
 22 |   data_param {
 23 |     source: "examples/imagenet/ilsvrc12_train_lmdb"
 24 |     batch_size: 32
 25 |     backend: LMDB
 26 |   }
 27 | }
 28 | layer {
 29 |   name: "data"
 30 |   type: "Data"
 31 |   top: "data"
 32 |   top: "label"
 33 |   include {
 34 |     phase: TEST
 35 |   }
 36 |   transform_param {
 37 |     crop_size: 227
 38 |     mean_value: 104
 39 |     mean_value: 117
 40 |     mean_value: 123
 41 |   }
 42 |   data_param {
 43 |     source: "examples/imagenet/ilsvrc12_val_lmdb"
 44 |     batch_size: 25 #not *iter_size
 45 |     backend: LMDB
 46 |   }
 47 | }
 48 | layer {
 49 |   name: "conv1"
 50 |   type: "Convolution"
 51 |   bottom: "data"
 52 |   top: "conv1"
 53 |   convolution_param {
 54 |     num_output: 64
 55 |     kernel_size: 3
 56 |     stride: 2
 57 |     weight_filler {
 58 |       type: "xavier"
 59 |     }
 60 |   }
 61 | }
 62 | layer {
 63 |   name: "relu_conv1"
 64 |   type: "ReLU"
 65 |   bottom: "conv1"
 66 |   top: "conv1"
 67 | }
 68 | layer {
 69 |   name: "pool1"
 70 |   type: "Pooling"
 71 |   bottom: "conv1"
 72 |   top: "pool1"
 73 |   pooling_param {
 74 |     pool: MAX
 75 |     kernel_size: 3
 76 |     stride: 2
 77 |   }
 78 | }
 79 | layer {
 80 |   name: "fire2/squeeze1x1"
 81 |   type: "Convolution"
 82 |   bottom: "pool1"
 83 |   top: "fire2/squeeze1x1"
 84 |   convolution_param {
 85 |     num_output: 16
 86 |     kernel_size: 1
 87 |     weight_filler {
 88 |       type: "xavier"
 89 |     }
 90 |   }
 91 | }
 92 | layer {
 93 |   name: "fire2/relu_squeeze1x1"
 94 |   type: "ReLU"
 95 |   bottom: "fire2/squeeze1x1"
 96 |   top: "fire2/squeeze1x1"
 97 | }
 98 | layer {
 99 |   name: "fire2/expand1x1"
100 |   type: "Convolution"
101 |   bottom: "fire2/squeeze1x1"
102 |   top: "fire2/expand1x1"
103 |   convolution_param {
104 |     num_output: 64
105 |     kernel_size: 1
106 |     weight_filler {
107 |       type: "xavier"
108 |     }
109 |   }
110 | }
111 | layer {
112 |   name: "fire2/relu_expand1x1"
113 |   type: "ReLU"
114 |   bottom: "fire2/expand1x1"
115 |   top: "fire2/expand1x1"
116 | }
117 | layer {
118 |   name: "fire2/expand3x3"
119 |   type: "Convolution"
120 |   bottom: "fire2/squeeze1x1"
121 |   top: "fire2/expand3x3"
122 |   convolution_param {
123 |     num_output: 64
124 |     pad: 1
125 |     kernel_size: 3
126 |     weight_filler {
127 |       type: "xavier"
128 |     }
129 |   }
130 | }
131 | layer {
132 |   name: "fire2/relu_expand3x3"
133 |   type: "ReLU"
134 |   bottom: "fire2/expand3x3"
135 |   top: "fire2/expand3x3"
136 | }
137 | layer {
138 |   name: "fire2/concat"
139 |   type: "Concat"
140 |   bottom: "fire2/expand1x1"
141 |   bottom: "fire2/expand3x3"
142 |   top: "fire2/concat"
143 | }
144 | layer {
145 |   name: "fire3/squeeze1x1"
146 |   type: "Convolution"
147 |   bottom: "fire2/concat"
148 |   top: "fire3/squeeze1x1"
149 |   convolution_param {
150 |     num_output: 16
151 |     kernel_size: 1
152 |     weight_filler {
153 |       type: "xavier"
154 |     }
155 |   }
156 | }
157 | layer {
158 |   name: "fire3/relu_squeeze1x1"
159 |   type: "ReLU"
160 |   bottom: "fire3/squeeze1x1"
161 |   top: "fire3/squeeze1x1"
162 | }
163 | layer {
164 |   name: "fire3/expand1x1"
165 |   type: "Convolution"
166 |   bottom: "fire3/squeeze1x1"
167 |   top: "fire3/expand1x1"
168 |   convolution_param {
169 |     num_output: 64
170 |     kernel_size: 1
171 |     weight_filler {
172 |       type: "xavier"
173 |     }
174 |   }
175 | }
176 | layer {
177 |   name: "fire3/relu_expand1x1"
178 |   type: "ReLU"
179 |   bottom: "fire3/expand1x1"
180 |   top: "fire3/expand1x1"
181 | }
182 | layer {
183 |   name: "fire3/expand3x3"
184 |   type: "Convolution"
185 |   bottom: "fire3/squeeze1x1"
186 |   top: "fire3/expand3x3"
187 |   convolution_param {
188 |     num_output: 64
189 |     pad: 1
190 |     kernel_size: 3
191 |     weight_filler {
192 |       type: "xavier"
193 |     }
194 |   }
195 | }
196 | layer {
197 |   name: "fire3/relu_expand3x3"
198 |   type: "ReLU"
199 |   bottom: "fire3/expand3x3"
200 |   top: "fire3/expand3x3"
201 | }
202 | layer {
203 |   name: "fire3/concat"
204 |   type: "Concat"
205 |   bottom: "fire3/expand1x1"
206 |   bottom: "fire3/expand3x3"
207 |   top: "fire3/concat"
208 | }
209 | layer {
210 |   name: "pool3"
211 |   type: "Pooling"
212 |   bottom: "fire3/concat"
213 |   top: "pool3"
214 |   pooling_param {
215 |     pool: MAX
216 |     kernel_size: 3
217 |     stride: 2
218 |   }
219 | }
220 | layer {
221 |   name: "fire4/squeeze1x1"
222 |   type: "Convolution"
223 |   bottom: "pool3"
224 |   top: "fire4/squeeze1x1"
225 |   convolution_param {
226 |     num_output: 32
227 |     kernel_size: 1
228 |     weight_filler {
229 |       type: "xavier"
230 |     }
231 |   }
232 | }
233 | layer {
234 |   name: "fire4/relu_squeeze1x1"
235 |   type: "ReLU"
236 |   bottom: "fire4/squeeze1x1"
237 |   top: "fire4/squeeze1x1"
238 | }
239 | layer {
240 |   name: "fire4/expand1x1"
241 |   type: "Convolution"
242 |   bottom: "fire4/squeeze1x1"
243 |   top: "fire4/expand1x1"
244 |   convolution_param {
245 |     num_output: 128
246 |     kernel_size: 1
247 |     weight_filler {
248 |       type: "xavier"
249 |     }
250 |   }
251 | }
252 | layer {
253 |   name: "fire4/relu_expand1x1"
254 |   type: "ReLU"
255 |   bottom: "fire4/expand1x1"
256 |   top: "fire4/expand1x1"
257 | }
258 | layer {
259 |   name: "fire4/expand3x3"
260 |   type: "Convolution"
261 |   bottom: "fire4/squeeze1x1"
262 |   top: "fire4/expand3x3"
263 |   convolution_param {
264 |     num_output: 128
265 |     pad: 1
266 |     kernel_size: 3
267 |     weight_filler {
268 |       type: "xavier"
269 |     }
270 |   }
271 | }
272 | layer {
273 |   name: "fire4/relu_expand3x3"
274 |   type: "ReLU"
275 |   bottom: "fire4/expand3x3"
276 |   top: "fire4/expand3x3"
277 | }
278 | layer {
279 |   name: "fire4/concat"
280 |   type: "Concat"
281 |   bottom: "fire4/expand1x1"
282 |   bottom: "fire4/expand3x3"
283 |   top: "fire4/concat"
284 | }
285 | layer {
286 |   name: "fire5/squeeze1x1"
287 |   type: "Convolution"
288 |   bottom: "fire4/concat"
289 |   top: "fire5/squeeze1x1"
290 |   convolution_param {
291 |     num_output: 32
292 |     kernel_size: 1
293 |     weight_filler {
294 |       type: "xavier"
295 |     }
296 |   }
297 | }
298 | layer {
299 |   name: "fire5/relu_squeeze1x1"
300 |   type: "ReLU"
301 |   bottom: "fire5/squeeze1x1"
302 |   top: "fire5/squeeze1x1"
303 | }
304 | layer {
305 |   name: "fire5/expand1x1"
306 |   type: "Convolution"
307 |   bottom: "fire5/squeeze1x1"
308 |   top: "fire5/expand1x1"
309 |   convolution_param {
310 |     num_output: 128
311 |     kernel_size: 1
312 |     weight_filler {
313 |       type: "xavier"
314 |     }
315 |   }
316 | }
317 | layer {
318 |   name: "fire5/relu_expand1x1"
319 |   type: "ReLU"
320 |   bottom: "fire5/expand1x1"
321 |   top: "fire5/expand1x1"
322 | }
323 | layer {
324 |   name: "fire5/expand3x3"
325 |   type: "Convolution"
326 |   bottom: "fire5/squeeze1x1"
327 |   top: "fire5/expand3x3"
328 |   convolution_param {
329 |     num_output: 128
330 |     pad: 1
331 |     kernel_size: 3
332 |     weight_filler {
333 |       type: "xavier"
334 |     }
335 |   }
336 | }
337 | layer {
338 |   name: "fire5/relu_expand3x3"
339 |   type: "ReLU"
340 |   bottom: "fire5/expand3x3"
341 |   top: "fire5/expand3x3"
342 | }
343 | layer {
344 |   name: "fire5/concat"
345 |   type: "Concat"
346 |   bottom: "fire5/expand1x1"
347 |   bottom: "fire5/expand3x3"
348 |   top: "fire5/concat"
349 | }
350 | layer {
351 |   name: "pool5"
352 |   type: "Pooling"
353 |   bottom: "fire5/concat"
354 |   top: "pool5"
355 |   pooling_param {
356 |     pool: MAX
357 |     kernel_size: 3
358 |     stride: 2
359 |   }
360 | }
361 | layer {
362 |   name: "fire6/squeeze1x1"
363 |   type: "Convolution"
364 |   bottom: "pool5"
365 |   top: "fire6/squeeze1x1"
366 |   convolution_param {
367 |     num_output: 48
368 |     kernel_size: 1
369 |     weight_filler {
370 |       type: "xavier"
371 |     }
372 |   }
373 | }
374 | layer {
375 |   name: "fire6/relu_squeeze1x1"
376 |   type: "ReLU"
377 |   bottom: "fire6/squeeze1x1"
378 |   top: "fire6/squeeze1x1"
379 | }
380 | layer {
381 |   name: "fire6/expand1x1"
382 |   type: "Convolution"
383 |   bottom: "fire6/squeeze1x1"
384 |   top: "fire6/expand1x1"
385 |   convolution_param {
386 |     num_output: 192
387 |     kernel_size: 1
388 |     weight_filler {
389 |       type: "xavier"
390 |     }
391 |   }
392 | }
393 | layer {
394 |   name: "fire6/relu_expand1x1"
395 |   type: "ReLU"
396 |   bottom: "fire6/expand1x1"
397 |   top: "fire6/expand1x1"
398 | }
399 | layer {
400 |   name: "fire6/expand3x3"
401 |   type: "Convolution"
402 |   bottom: "fire6/squeeze1x1"
403 |   top: "fire6/expand3x3"
404 |   convolution_param {
405 |     num_output: 192
406 |     pad: 1
407 |     kernel_size: 3
408 |     weight_filler {
409 |       type: "xavier"
410 |     }
411 |   }
412 | }
413 | layer {
414 |   name: "fire6/relu_expand3x3"
415 |   type: "ReLU"
416 |   bottom: "fire6/expand3x3"
417 |   top: "fire6/expand3x3"
418 | }
419 | layer {
420 |   name: "fire6/concat"
421 |   type: "Concat"
422 |   bottom: "fire6/expand1x1"
423 |   bottom: "fire6/expand3x3"
424 |   top: "fire6/concat"
425 | }
426 | layer {
427 |   name: "fire7/squeeze1x1"
428 |   type: "Convolution"
429 |   bottom: "fire6/concat"
430 |   top: "fire7/squeeze1x1"
431 |   convolution_param {
432 |     num_output: 48
433 |     kernel_size: 1
434 |     weight_filler {
435 |       type: "xavier"
436 |     }
437 |   }
438 | }
439 | layer {
440 |   name: "fire7/relu_squeeze1x1"
441 |   type: "ReLU"
442 |   bottom: "fire7/squeeze1x1"
443 |   top: "fire7/squeeze1x1"
444 | }
445 | layer {
446 |   name: "fire7/expand1x1"
447 |   type: "Convolution"
448 |   bottom: "fire7/squeeze1x1"
449 |   top: "fire7/expand1x1"
450 |   convolution_param {
451 |     num_output: 192
452 |     kernel_size: 1
453 |     weight_filler {
454 |       type: "xavier"
455 |     }
456 |   }
457 | }
458 | layer {
459 |   name: "fire7/relu_expand1x1"
460 |   type: "ReLU"
461 |   bottom: "fire7/expand1x1"
462 |   top: "fire7/expand1x1"
463 | }
464 | layer {
465 |   name: "fire7/expand3x3"
466 |   type: "Convolution"
467 |   bottom: "fire7/squeeze1x1"
468 |   top: "fire7/expand3x3"
469 |   convolution_param {
470 |     num_output: 192
471 |     pad: 1
472 |     kernel_size: 3
473 |     weight_filler {
474 |       type: "xavier"
475 |     }
476 |   }
477 | }
478 | layer {
479 |   name: "fire7/relu_expand3x3"
480 |   type: "ReLU"
481 |   bottom: "fire7/expand3x3"
482 |   top: "fire7/expand3x3"
483 | }
484 | layer {
485 |   name: "fire7/concat"
486 |   type: "Concat"
487 |   bottom: "fire7/expand1x1"
488 |   bottom: "fire7/expand3x3"
489 |   top: "fire7/concat"
490 | }
491 | layer {
492 |   name: "fire8/squeeze1x1"
493 |   type: "Convolution"
494 |   bottom: "fire7/concat"
495 |   top: "fire8/squeeze1x1"
496 |   convolution_param {
497 |     num_output: 64
498 |     kernel_size: 1
499 |     weight_filler {
500 |       type: "xavier"
501 |     }
502 |   }
503 | }
504 | layer {
505 |   name: "fire8/relu_squeeze1x1"
506 |   type: "ReLU"
507 |   bottom: "fire8/squeeze1x1"
508 |   top: "fire8/squeeze1x1"
509 | }
510 | layer {
511 |   name: "fire8/expand1x1"
512 |   type: "Convolution"
513 |   bottom: "fire8/squeeze1x1"
514 |   top: "fire8/expand1x1"
515 |   convolution_param {
516 |     num_output: 256
517 |     kernel_size: 1
518 |     weight_filler {
519 |       type: "xavier"
520 |     }
521 |   }
522 | }
523 | layer {
524 |   name: "fire8/relu_expand1x1"
525 |   type: "ReLU"
526 |   bottom: "fire8/expand1x1"
527 |   top: "fire8/expand1x1"
528 | }
529 | layer {
530 |   name: "fire8/expand3x3"
531 |   type: "Convolution"
532 |   bottom: "fire8/squeeze1x1"
533 |   top: "fire8/expand3x3"
534 |   convolution_param {
535 |     num_output: 256
536 |     pad: 1
537 |     kernel_size: 3
538 |     weight_filler {
539 |       type: "xavier"
540 |     }
541 |   }
542 | }
543 | layer {
544 |   name: "fire8/relu_expand3x3"
545 |   type: "ReLU"
546 |   bottom: "fire8/expand3x3"
547 |   top: "fire8/expand3x3"
548 | }
549 | layer {
550 |   name: "fire8/concat"
551 |   type: "Concat"
552 |   bottom: "fire8/expand1x1"
553 |   bottom: "fire8/expand3x3"
554 |   top: "fire8/concat"
555 | }
556 | layer {
557 |   name: "fire9/squeeze1x1"
558 |   type: "Convolution"
559 |   bottom: "fire8/concat"
560 |   top: "fire9/squeeze1x1"
561 |   convolution_param {
562 |     num_output: 64
563 |     kernel_size: 1
564 |     weight_filler {
565 |       type: "xavier"
566 |     }
567 |   }
568 | }
569 | layer {
570 |   name: "fire9/relu_squeeze1x1"
571 |   type: "ReLU"
572 |   bottom: "fire9/squeeze1x1"
573 |   top: "fire9/squeeze1x1"
574 | }
575 | layer {
576 |   name: "fire9/expand1x1"
577 |   type: "Convolution"
578 |   bottom: "fire9/squeeze1x1"
579 |   top: "fire9/expand1x1"
580 |   convolution_param {
581 |     num_output: 256
582 |     kernel_size: 1
583 |     weight_filler {
584 |       type: "xavier"
585 |     }
586 |   }
587 | }
588 | layer {
589 |   name: "fire9/relu_expand1x1"
590 |   type: "ReLU"
591 |   bottom: "fire9/expand1x1"
592 |   top: "fire9/expand1x1"
593 | }
594 | layer {
595 |   name: "fire9/expand3x3"
596 |   type: "Convolution"
597 |   bottom: "fire9/squeeze1x1"
598 |   top: "fire9/expand3x3"
599 |   convolution_param {
600 |     num_output: 256
601 |     pad: 1
602 |     kernel_size: 3
603 |     weight_filler {
604 |       type: "xavier"
605 |     }
606 |   }
607 | }
608 | layer {
609 |   name: "fire9/relu_expand3x3"
610 |   type: "ReLU"
611 |   bottom: "fire9/expand3x3"
612 |   top: "fire9/expand3x3"
613 | }
614 | layer {
615 |   name: "fire9/concat"
616 |   type: "Concat"
617 |   bottom: "fire9/expand1x1"
618 |   bottom: "fire9/expand3x3"
619 |   top: "fire9/concat"
620 | }
621 | layer {
622 |   name: "drop9"
623 |   type: "Dropout"
624 |   bottom: "fire9/concat"
625 |   top: "fire9/concat"
626 |   dropout_param {
627 |     dropout_ratio: 0.5
628 |   }
629 | }
630 | layer {
631 |   name: "conv10"
632 |   type: "Convolution"
633 |   bottom: "fire9/concat"
634 |   top: "conv10"
635 |   convolution_param {
636 |     num_output: 1000
637 |     kernel_size: 1
638 |     weight_filler {
639 |       type: "gaussian"
640 |       mean: 0.0
641 |       std: 0.01
642 |     }
643 |   }
644 | }
645 | layer {
646 |   name: "relu_conv10"
647 |   type: "ReLU"
648 |   bottom: "conv10"
649 |   top: "conv10"
650 | }
651 | layer {
652 |   name: "pool10"
653 |   type: "Pooling"
654 |   bottom: "conv10"
655 |   top: "pool10"
656 |   pooling_param {
657 |     pool: AVE
658 |     global_pooling: true
659 |   }
660 | }
661 | layer {
662 |   name: "loss"
663 |   type: "SoftmaxWithLoss"
664 |   bottom: "pool10"
665 |   bottom: "label"
666 |   top: "loss"
667 |   #include {
668 |   #  phase: TRAIN
669 |   #}
670 | }
671 | layer {
672 |   name: "accuracy"
673 |   type: "Accuracy"
674 |   bottom: "pool10"
675 |   bottom: "label"
676 |   top: "accuracy"
677 |   #include {
678 |   #  phase: TEST
679 |   #}
680 | }
681 | layer {
682 |   name: "accuracy_top5"
683 |   type: "Accuracy"
684 |   bottom: "pool10"
685 |   bottom: "label"
686 |   top: "accuracy_top5"
687 |   #include {
688 |   #  phase: TEST
689 |   #}
690 |   accuracy_param {
691 |     top_k: 5
692 |   }
693 | }
694 | 


--------------------------------------------------------------------------------