1 /******************************************************************************/\r
2 /*!\r
3 * \file testCNNSimpleTraining.c\r
4 */\r
5 /* Copyright (C) 2015 Texas Instruments Incorporated - http://www.ti.com/\r
6 *\r
7 * Redistribution and use in source and binary forms, with or without\r
8 * modification, are permitted provided that the following conditions\r
9 * are met:\r
10 *\r
11 * Redistributions of source code must retain the above copyright\r
12 * notice, this list of conditions and the following disclaimer.\r
13 *\r
14 * Redistributions in binary form must reproduce the above copyright\r
15 * notice, this list of conditions and the following disclaimer in the\r
16 * documentation and/or other materials provided with the\r
17 * distribution.\r
18 *\r
19 * Neither the name of Texas Instruments Incorporated nor the names of\r
20 * its contributors may be used to endorse or promote products derived\r
21 * from this software without specific prior written permission.\r
22 *\r
23 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS\r
24 * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT\r
25 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR\r
26 * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT\r
27 * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,\r
28 * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT\r
29 * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,\r
30 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY\r
31 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT\r
32 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE\r
33 * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.\r
34 *\r
35 ******************************************************************************/\r
36 \r
37 \r
38 /*******************************************************************************\r
39 *\r
40 * INCLUDES\r
41 *\r
42 ******************************************************************************/\r
43 \r
44 #include "testCNN.h"\r
45 \r
46 \r
47 /*******************************************************************************\r
48 *\r
49 * DEFINES\r
50 *\r
51 ******************************************************************************/\r
52 \r
53 #define DATABASE_PATH "./src/database/mnist"\r
54 #define BATCH_SIZE 100 // cnn batch size\r
55 #define MAX_BATCH_SIZE 100 // cnn max batch size\r
56 #define BATCH_UPDATE 100 // cnn batch update size (must be a multiple of batch size)\r
57 #define TEST_NUM 10000\r
58 #define TRAIN_NUM 60000\r
59 #define IMAGE_ROW 28\r
60 #define IMAGE_COL 28\r
61 #define IMAGE_CHANNEL 1\r
62 #define TOP_N 1\r
63 \r
64 \r
65 /******************************************************************************/\r
66 /*!\r
67 * \ingroup testCNN\r
68 * \brief Simple training function test\r
69 * \return Error code\r
70 */\r
71 /******************************************************************************/\r
72 \r
73 int testCNNSimpleTraining()\r
74 {\r
75 int i;\r
76 int dim;\r
77 int classifyNum;\r
78 float classifyPercent;\r
79 int label;\r
80 struct timespec startTime;\r
81 struct timespec endTime;\r
82 long trainingTime;\r
83 long testingTime;\r
84 int err;\r
85 long mem;\r
86 int batchNum;\r
87 int batchSize;\r
88 timlUtilImageSet training;\r
89 timlUtilImageSet testing;\r
90 timlCNNInputParams inputParams;\r
91 timlCNNTrainingParams trainingParams;\r
92 timlConvNeuralNetwork *cnn;\r
93 timlCNNLayer *layer;\r
94 \r
95 dim = IMAGE_ROW*IMAGE_COL*IMAGE_CHANNEL;\r
96 classifyNum = 0;\r
97 err = 0;\r
98 \r
99 setbuf(stdout, NULL); // do not buffer the console output\r
100 \r
101 printf("[Test] CNN simple training\n");\r
102 printf("1. Build up the CNN\n");\r
103 trainingParams = timlCNNTrainingParamsDefault();\r
104 trainingParams.batchSize = BATCH_SIZE;\r
105 trainingParams.maxBatchSize = MAX_BATCH_SIZE;\r
106 trainingParams.batchUpdate = BATCH_UPDATE;\r
107 trainingParams.learningRate = 0.1;\r
108 cnn = timlCNNCreateConvNeuralNetwork(trainingParams);\r
109 inputParams = timlCNNInputParamsDefault();\r
110 inputParams.scale = 1.0/256.0;\r
111 timlCNNAddInputLayer(cnn, IMAGE_ROW, IMAGE_COL, IMAGE_CHANNEL, inputParams); // input layer\r
112 timlCNNAddConvLayer(cnn, 5, 5, 1, 1, 6, timlCNNConvParamsDefault()); // conv layer\r
113 timlCNNAddNonlinearLayer(cnn, Util_Relu); // relu layer\r
114 timlCNNAddPoolingLayer(cnn, 4, 4, 4, 4, CNN_MaxPooling, timlCNNPoolingParamsDefault()); // max pooling layer\r
115 timlCNNAddNormLayer(cnn, timlCNNNormParamsDefault()); // norm layer\r
116 timlCNNAddDropoutLayer(cnn, 0.5); // dropout layer\r
117 timlCNNAddLinearLayer(cnn, 10, timlCNNLinearParamsDefault()); // linear layer\r
118 timlCNNAddSoftmaxCostLayer(cnn); // softmax cost layer\r
119 timlCNNInitialize(cnn);\r
120 timlCNNReset(cnn);\r
121 timlCNNPrint(cnn);\r
122 \r
123 printf("2. Load the MNIST database\n");\r
124 err = timlUtilReadMNIST(DATABASE_PATH, &training, &testing);\r
125 if (err) {\r
126 printf("MNIST database reading error\n");\r
127 }\r
128 \r
129 // training\r
130 // timlCNNSetBatchSize(cnn, 1);\r
131 printf("3. Start training\n");\r
132 timlCNNSetMode(cnn, Util_Train);\r
133 batchNum = TRAIN_NUM/cnn->params.batchUpdate;\r
134 clock_gettime(CLOCK_REALTIME, &startTime);\r
135 for (i = 0; i < batchNum; i++) {\r
136 printf("Batch id = %d, ", i);\r
137 timlCNNSupervisedTrainingWithLabel(cnn, training.data + i*cnn->params.batchUpdate*dim, IMAGE_ROW, IMAGE_COL, IMAGE_CHANNEL, training.label + i*cnn->params.batchUpdate, 1, 1, cnn->params.batchUpdate);\r
138 }\r
139 clock_gettime(CLOCK_REALTIME, &endTime);\r
140 trainingTime = timlUtilDiffTime(startTime, endTime);\r
141 printf("Training time = %.3f s.\n", trainingTime/1000000.0);\r
142 \r
143 // testing\r
144 printf("4. Start testing\n");\r
145 // remove softmaxCost layer and add accuracy layer\r
146 timlCNNDeleteLayer(cnn->tail);\r
147 timlCNNAddAccuracyLayer(cnn, TOP_N);\r
148 timlCNNAccuracyInitialize(cnn->tail);\r
149 timlCNNSetMode(cnn, Util_Test);\r
150 \r
151 timlCNNSetBatchSize(cnn, 100);\r
152 clock_gettime(CLOCK_REALTIME, &startTime);\r
153 timlCNNClassifyAccuracy(cnn, testing.data, IMAGE_ROW, IMAGE_COL, IMAGE_CHANNEL, testing.label, 1, 1, testing.num, &classifyNum);\r
154 clock_gettime(CLOCK_REALTIME, &endTime);\r
155 testingTime = timlUtilDiffTime(startTime, endTime);\r
156 classifyPercent = (float)classifyNum/(float)testing.num;\r
157 printf("Testing time = %.3f s.\nSuccess percent = %.3f %%.\n", testingTime/1000000.0, classifyPercent*100.0);\r
158 \r
159 printf("5. Clean up\n");\r
160 free(training.data);\r
161 free(training.label);\r
162 free(training.mean);\r
163 free(testing.data);\r
164 free(testing.label);\r
165 timlCNNDelete(cnn);\r
166 \r
167 return err;\r
168 }\r