Trying to do Image segmentation by neural networks, but mini-batch accuracy and mini-batch loss are fluctuating

Hi,

I need some help with the problem I mentioned in the summarized question. I have images about engraved surfaces, I would like to train the network to segment the engraved and the not engraved areas on them. The learning process goes well until the 11th iteration, where it starts fluctuating. I have 65 [200x200] training images (not too much, but should be enough to get a partly "smart" network at the end). After the 100 epochs I open 2 figures (1 picture from the training images and 1 completely new picture).

Most of the time as a result I get every pixels classified as "non_engraved". These times the mini-batch accuracy remains constant after earned the 72%.

This time (randomly) the mini-batch accuracy went up to 86% (but still with fluctuating), and got a better result, but it is still disappointment:

|========================================================================================|

| Epoch | Iteration | Time Elapsed | Mini-batch | Mini-batch | Base Learning |

| | | (hh:mm:ss) | Accuracy | Loss | Rate |

|========================================================================================|

| 1 | 1 | 00:00:01 | 55.51% | 0.8153 | 0.0010 |

| 1 | 2 | 00:00:02 | 58.80% | 0.6984 | 0.0010 |

| 1 | 3 | 00:00:03 | 62.44% | 0.6768 | 0.0010 |

| 2 | 4 | 00:00:04 | 65.03% | 0.6759 | 0.0010 |

| 2 | 5 | 00:00:05 | 71.44% | 0.6360 | 0.0010 |

| 2 | 6 | 00:00:06 | 69.98% | 0.6468 | 0.0010 |

| 3 | 7 | 00:00:07 | 69.64% | 0.6484 | 0.0010 |

| 3 | 8 | 00:00:08 | 75.74% | 0.5924 | 0.0010 |

| 3 | 9 | 00:00:09 | 71.98% | 0.6131 | 0.0010 |

| 4 | 10 | 00:00:10 | 70.33% | 0.6183 | 0.0010 |

| 4 | 11 | 00:00:11 | 76.08% | 0.5759 | 0.0010 |

| 4 | 12 | 00:00:11 | 72.11% | 0.5857 | 0.0010 |

| 5 | 13 | 00:00:13 | 70.39% | 0.5918 | 0.0010 |

| 5 | 14 | 00:00:14 | 76.10% | 0.5708 | 0.0010 |

| 5 | 15 | 00:00:14 | 72.12% | 0.5690 | 0.0010 |

| 6 | 16 | 00:00:16 | 70.40% | 0.5777 | 0.0010 |

| 6 | 17 | 00:00:16 | 76.10% | 0.5552 | 0.0010 |

| 6 | 18 | 00:00:17 | 72.12% | 0.5680 | 0.0010 |

| 7 | 19 | 00:00:19 | 70.40% | 0.5641 | 0.0010 |

| 7 | 20 | 00:00:20 | 76.10% | 0.5485 | 0.0010 |

| 7 | 21 | 00:00:21 | 72.12% | 0.5665 | 0.0010 |

| 8 | 22 | 00:00:22 | 70.40% | 0.5532 | 0.0010 |

| 8 | 23 | 00:00:23 | 76.10% | 0.5499 | 0.0010 |

| 8 | 24 | 00:00:23 | 72.12% | 0.5551 | 0.0010 |

| 9 | 25 | 00:00:25 | 70.40% | 0.5501 | 0.0010 |

| 9 | 26 | 00:00:26 | 76.10% | 0.5532 | 0.0010 |

| 9 | 27 | 00:00:26 | 72.12% | 0.5501 | 0.0010 |

| 10 | 28 | 00:00:28 | 70.40% | 0.5418 | 0.0010 |

| 10 | 29 | 00:00:29 | 76.10% | 0.5513 | 0.0010 |

| 10 | 30 | 00:00:30 | 72.12% | 0.5523 | 0.0010 |

| 11 | 31 | 00:00:31 | 70.40% | 0.5391 | 0.0010 |

| 11 | 32 | 00:00:32 | 76.10% | 0.5433 | 0.0010 |

| 11 | 33 | 00:00:32 | 72.12% | 0.5516 | 0.0010 |

| 12 | 34 | 00:00:34 | 70.40% | 0.5398 | 0.0010 |

| 12 | 35 | 00:00:36 | 76.10% | 0.5415 | 0.0010 |

| 12 | 36 | 00:00:38 | 72.12% | 0.5480 | 0.0010 |

| 13 | 37 | 00:00:40 | 70.40% | 0.5355 | 0.0010 |

| 13 | 38 | 00:00:41 | 76.10% | 0.5433 | 0.0010 |

| 13 | 39 | 00:00:42 | 72.12% | 0.5440 | 0.0010 |

| 14 | 40 | 00:00:43 | 70.40% | 0.5342 | 0.0010 |

| 14 | 41 | 00:00:44 | 76.10% | 0.5404 | 0.0010 |

| 14 | 42 | 00:00:45 | 72.12% | 0.5419 | 0.0010 |

| 15 | 43 | 00:00:48 | 70.40% | 0.5331 | 0.0010 |

| 15 | 44 | 00:00:50 | 76.10% | 0.5374 | 0.0010 |

| 15 | 45 | 00:00:51 | 72.12% | 0.5414 | 0.0010 |

| 16 | 46 | 00:00:53 | 70.40% | 0.5272 | 0.0010 |

| 16 | 47 | 00:00:54 | 76.10% | 0.5370 | 0.0010 |

| 16 | 48 | 00:00:55 | 72.12% | 0.5401 | 0.0010 |

| 17 | 49 | 00:00:56 | 70.40% | 0.5247 | 0.0010 |

| 17 | 50 | 00:00:57 | 76.10% | 0.5338 | 0.0010 |

| 17 | 51 | 00:00:58 | 72.12% | 0.5370 | 0.0010 |

| 18 | 52 | 00:00:59 | 70.40% | 0.5235 | 0.0010 |

| 18 | 53 | 00:01:00 | 76.10% | 0.5324 | 0.0010 |

| 18 | 54 | 00:01:00 | 72.12% | 0.5348 | 0.0010 |

| 19 | 55 | 00:01:02 | 70.40% | 0.5191 | 0.0010 |

| 19 | 56 | 00:01:02 | 76.10% | 0.5306 | 0.0010 |

| 19 | 57 | 00:01:03 | 72.12% | 0.5335 | 0.0010 |

| 20 | 58 | 00:01:04 | 70.40% | 0.5172 | 0.0010 |

| 20 | 59 | 00:01:05 | 76.10% | 0.5272 | 0.0010 |

| 20 | 60 | 00:01:06 | 72.12% | 0.5320 | 0.0010 |

| 21 | 61 | 00:01:07 | 70.40% | 0.5145 | 0.0010 |

| 21 | 62 | 00:01:08 | 76.10% | 0.5258 | 0.0010 |

| 21 | 63 | 00:01:09 | 72.12% | 0.5303 | 0.0010 |

| 22 | 64 | 00:01:10 | 70.40% | 0.5118 | 0.0010 |

| 22 | 65 | 00:01:11 | 76.10% | 0.5243 | 0.0010 |

| 22 | 66 | 00:01:12 | 72.13% | 0.5281 | 0.0010 |

| 23 | 67 | 00:01:13 | 70.40% | 0.5089 | 0.0010 |

| 23 | 68 | 00:01:14 | 76.10% | 0.5230 | 0.0010 |

| 23 | 69 | 00:01:15 | 72.13% | 0.5269 | 0.0010 |

| 24 | 70 | 00:01:16 | 70.40% | 0.5070 | 0.0010 |

| 24 | 71 | 00:01:17 | 76.10% | 0.5207 | 0.0010 |

| 24 | 72 | 00:01:18 | 72.13% | 0.5252 | 0.0010 |

| 25 | 73 | 00:01:19 | 70.41% | 0.5046 | 0.0010 |

| 25 | 74 | 00:01:20 | 76.10% | 0.5200 | 0.0010 |

| 25 | 75 | 00:01:21 | 72.14% | 0.5235 | 0.0010 |

| 26 | 76 | 00:01:23 | 70.41% | 0.5023 | 0.0010 |

| 26 | 77 | 00:01:24 | 76.10% | 0.5191 | 0.0010 |

| 26 | 78 | 00:01:25 | 72.14% | 0.5208 | 0.0010 |

| 27 | 79 | 00:01:26 | 70.42% | 0.5005 | 0.0010 |

| 27 | 80 | 00:01:27 | 76.10% | 0.5172 | 0.0010 |

| 27 | 81 | 00:01:28 | 72.15% | 0.5203 | 0.0010 |

| 28 | 82 | 00:01:29 | 70.44% | 0.4983 | 0.0010 |

| 28 | 83 | 00:01:30 | 76.11% | 0.5146 | 0.0010 |

| 28 | 84 | 00:01:31 | 72.16% | 0.5185 | 0.0010 |

| 29 | 85 | 00:01:33 | 70.48% | 0.4962 | 0.0010 |

| 29 | 86 | 00:01:33 | 76.11% | 0.5140 | 0.0010 |

| 29 | 87 | 00:01:34 | 72.23% | 0.5165 | 0.0010 |

| 30 | 88 | 00:01:36 | 76.55% | 0.4941 | 0.0010 |

| 30 | 89 | 00:01:37 | 77.57% | 0.5121 | 0.0010 |

| 30 | 90 | 00:01:37 | 75.40% | 0.5153 | 0.0010 |

| 31 | 91 | 00:01:39 | 77.27% | 0.4914 | 0.0010 |

| 31 | 92 | 00:01:40 | 77.67% | 0.5107 | 0.0010 |

| 31 | 93 | 00:01:40 | 75.66% | 0.5138 | 0.0010 |

| 32 | 94 | 00:01:42 | 77.84% | 0.4900 | 0.0010 |

| 32 | 95 | 00:01:43 | 77.74% | 0.5087 | 0.0010 |

| 32 | 96 | 00:01:43 | 75.80% | 0.5120 | 0.0010 |

| 33 | 97 | 00:01:45 | 78.06% | 0.4873 | 0.0010 |

| 33 | 98 | 00:01:46 | 77.77% | 0.5072 | 0.0010 |

| 33 | 99 | 00:01:47 | 75.97% | 0.5103 | 0.0010 |

| 34 | 100 | 00:01:48 | 78.48% | 0.4858 | 0.0010 |

| 34 | 101 | 00:01:49 | 77.82% | 0.5054 | 0.0010 |

| 34 | 102 | 00:01:50 | 76.06% | 0.5091 | 0.0010 |

| 35 | 103 | 00:01:51 | 78.72% | 0.4829 | 0.0010 |

| 35 | 104 | 00:01:52 | 77.88% | 0.5037 | 0.0010 |

| 35 | 105 | 00:01:53 | 76.16% | 0.5070 | 0.0010 |

| 36 | 106 | 00:01:54 | 79.01% | 0.4810 | 0.0010 |

| 36 | 107 | 00:01:55 | 77.89% | 0.5023 | 0.0010 |

| 36 | 108 | 00:01:56 | 76.23% | 0.5054 | 0.0010 |

| 37 | 109 | 00:01:57 | 79.15% | 0.4783 | 0.0010 |

| 37 | 110 | 00:01:58 | 77.88% | 0.5000 | 0.0010 |

| 37 | 111 | 00:01:59 | 76.32% | 0.5038 | 0.0010 |

| 38 | 112 | 00:02:00 | 79.39% | 0.4762 | 0.0010 |

| 38 | 113 | 00:02:01 | 77.91% | 0.4983 | 0.0010 |

| 38 | 114 | 00:02:02 | 76.42% | 0.5019 | 0.0010 |

| 39 | 115 | 00:02:03 | 79.60% | 0.4736 | 0.0010 |

| 39 | 116 | 00:02:04 | 77.85% | 0.4967 | 0.0010 |

| 39 | 117 | 00:02:05 | 76.49% | 0.4998 | 0.0010 |

| 40 | 118 | 00:02:06 | 79.82% | 0.4715 | 0.0010 |

| 40 | 119 | 00:02:07 | 77.91% | 0.4949 | 0.0010 |

| 40 | 120 | 00:02:08 | 76.59% | 0.4972 | 0.0010 |

| 41 | 121 | 00:02:10 | 79.94% | 0.4682 | 0.0010 |

| 41 | 122 | 00:02:10 | 77.91% | 0.4935 | 0.0010 |

| 41 | 123 | 00:02:11 | 76.71% | 0.4954 | 0.0010 |

| 42 | 124 | 00:02:13 | 80.26% | 0.4667 | 0.0010 |

| 42 | 125 | 00:02:13 | 77.98% | 0.4914 | 0.0010 |

| 42 | 126 | 00:02:14 | 76.75% | 0.4936 | 0.0010 |

| 43 | 127 | 00:02:16 | 80.38% | 0.4635 | 0.0010 |

| 43 | 128 | 00:02:16 | 77.99% | 0.4894 | 0.0010 |

| 43 | 129 | 00:02:17 | 76.85% | 0.4909 | 0.0010 |

| 44 | 130 | 00:02:19 | 80.55% | 0.4612 | 0.0010 |

| 44 | 131 | 00:02:20 | 78.06% | 0.4873 | 0.0010 |

| 44 | 132 | 00:02:21 | 76.92% | 0.4892 | 0.0010 |

| 45 | 133 | 00:02:23 | 80.75% | 0.4592 | 0.0010 |

| 45 | 134 | 00:02:24 | 78.06% | 0.4852 | 0.0010 |

| 45 | 135 | 00:02:25 | 76.98% | 0.4864 | 0.0010 |

| 46 | 136 | 00:02:26 | 80.92% | 0.4567 | 0.0010 |

| 46 | 137 | 00:02:27 | 78.21% | 0.4829 | 0.0010 |

| 46 | 138 | 00:02:29 | 77.05% | 0.4840 | 0.0010 |

| 47 | 139 | 00:02:31 | 81.00% | 0.4527 | 0.0010 |

| 47 | 140 | 00:02:34 | 78.21% | 0.4809 | 0.0010 |

| 47 | 141 | 00:02:35 | 77.19% | 0.4825 | 0.0010 |

| 48 | 142 | 00:02:38 | 81.24% | 0.4510 | 0.0010 |

| 48 | 143 | 00:02:39 | 78.25% | 0.4798 | 0.0010 |

| 48 | 144 | 00:02:41 | 77.19% | 0.4797 | 0.0010 |

| 49 | 145 | 00:02:43 | 81.45% | 0.4501 | 0.0010 |

| 49 | 146 | 00:02:45 | 78.29% | 0.4764 | 0.0010 |

| 49 | 147 | 00:02:46 | 77.21% | 0.4775 | 0.0010 |

| 50 | 148 | 00:02:47 | 81.52% | 0.4455 | 0.0010 |

| 50 | 149 | 00:02:48 | 78.44% | 0.4732 | 0.0010 |

| 50 | 150 | 00:02:49 | 77.41% | 0.4751 | 0.0010 |

| 51 | 151 | 00:02:51 | 81.52% | 0.4416 | 0.0010 |

| 51 | 152 | 00:02:52 | 78.48% | 0.4746 | 0.0010 |

| 51 | 153 | 00:02:53 | 77.49% | 0.4719 | 0.0010 |

| 52 | 154 | 00:02:55 | 81.84% | 0.4410 | 0.0010 |

| 52 | 155 | 00:02:56 | 78.41% | 0.4721 | 0.0010 |

| 52 | 156 | 00:02:58 | 77.52% | 0.4705 | 0.0010 |

| 53 | 157 | 00:03:00 | 82.06% | 0.4391 | 0.0010 |

| 53 | 158 | 00:03:01 | 78.52% | 0.4694 | 0.0010 |

| 53 | 159 | 00:03:02 | 77.51% | 0.4685 | 0.0010 |

| 54 | 160 | 00:03:03 | 82.13% | 0.4365 | 0.0010 |

| 54 | 161 | 00:03:05 | 78.76% | 0.4656 | 0.0010 |

| 54 | 162 | 00:03:07 | 77.62% | 0.4647 | 0.0010 |

| 55 | 163 | 00:03:10 | 82.09% | 0.4314 | 0.0010 |

| 55 | 164 | 00:03:11 | 78.79% | 0.4636 | 0.0010 |

| 55 | 165 | 00:03:13 | 77.90% | 0.4615 | 0.0010 |

| 56 | 166 | 00:03:18 | 82.22% | 0.4271 | 0.0010 |

| 56 | 167 | 00:03:20 | 78.76% | 0.4654 | 0.0010 |

| 56 | 168 | 00:03:21 | 78.09% | 0.4597 | 0.0010 |

| 57 | 169 | 00:03:22 | 82.51% | 0.4267 | 0.0010 |

| 57 | 170 | 00:03:23 | 78.67% | 0.4648 | 0.0010 |

| 57 | 171 | 00:03:24 | 78.02% | 0.4563 | 0.0010 |

| 58 | 172 | 00:03:25 | 82.89% | 0.4300 | 0.0010 |

| 58 | 173 | 00:03:26 | 78.88% | 0.4589 | 0.0010 |

| 58 | 174 | 00:03:27 | 77.80% | 0.4589 | 0.0010 |

| 59 | 175 | 00:03:29 | 82.95% | 0.4250 | 0.0010 |

| 59 | 176 | 00:03:30 | 79.43% | 0.4560 | 0.0010 |

| 59 | 177 | 00:03:31 | 77.98% | 0.4537 | 0.0010 |

| 60 | 178 | 00:03:32 | 82.72% | 0.4166 | 0.0010 |

| 60 | 179 | 00:03:33 | 79.50% | 0.4528 | 0.0010 |

| 60 | 180 | 00:03:34 | 78.56% | 0.4487 | 0.0010 |

| 61 | 181 | 00:03:35 | 82.69% | 0.4125 | 0.0010 |

| 61 | 182 | 00:03:36 | 79.26% | 0.4542 | 0.0010 |

| 61 | 183 | 00:03:37 | 78.76% | 0.4465 | 0.0010 |

| 62 | 184 | 00:03:38 | 83.15% | 0.4131 | 0.0010 |

| 62 | 185 | 00:03:39 | 79.05% | 0.4587 | 0.0010 |

| 62 | 186 | 00:03:40 | 78.57% | 0.4420 | 0.0010 |

| 63 | 187 | 00:03:41 | 83.58% | 0.4205 | 0.0010 |

| 63 | 188 | 00:03:43 | 79.55% | 0.4484 | 0.0010 |

| 63 | 189 | 00:03:44 | 77.99% | 0.4504 | 0.0010 |

| 64 | 190 | 00:03:45 | 83.71% | 0.4112 | 0.0010 |

| 64 | 191 | 00:03:46 | 80.35% | 0.4487 | 0.0010 |

| 64 | 192 | 00:03:47 | 78.61% | 0.4414 | 0.0010 |

| 65 | 193 | 00:03:49 | 83.13% | 0.4027 | 0.0010 |

| 65 | 194 | 00:03:50 | 80.29% | 0.4426 | 0.0010 |

| 65 | 195 | 00:03:51 | 79.31% | 0.4408 | 0.0010 |

| 66 | 196 | 00:03:53 | 83.19% | 0.3992 | 0.0010 |

| 66 | 197 | 00:03:54 | 79.61% | 0.4496 | 0.0010 |

| 66 | 198 | 00:03:55 | 79.40% | 0.4303 | 0.0010 |

| 67 | 199 | 00:03:57 | 83.99% | 0.4067 | 0.0010 |

| 67 | 200 | 00:03:58 | 79.79% | 0.4452 | 0.0010 |

| 67 | 201 | 00:03:59 | 78.64% | 0.4356 | 0.0010 |

| 68 | 202 | 00:04:00 | 84.24% | 0.4040 | 0.0010 |

| 68 | 203 | 00:04:01 | 80.83% | 0.4390 | 0.0010 |

| 68 | 204 | 00:04:03 | 78.87% | 0.4353 | 0.0010 |

| 69 | 205 | 00:04:04 | 83.85% | 0.3901 | 0.0010 |

| 69 | 206 | 00:04:05 | 80.93% | 0.4380 | 0.0010 |

| 69 | 207 | 00:04:06 | 79.84% | 0.4258 | 0.0010 |

| 70 | 208 | 00:04:07 | 83.60% | 0.3899 | 0.0010 |

| 70 | 209 | 00:04:08 | 80.33% | 0.4379 | 0.0010 |

| 70 | 210 | 00:04:09 | 80.04% | 0.4241 | 0.0010 |

| 71 | 211 | 00:04:10 | 84.47% | 0.3914 | 0.0010 |

| 71 | 212 | 00:04:11 | 80.15% | 0.4392 | 0.0010 |

| 71 | 213 | 00:04:12 | 79.32% | 0.4234 | 0.0010 |

| 72 | 214 | 00:04:14 | 84.72% | 0.3963 | 0.0010 |

| 72 | 215 | 00:04:15 | 81.34% | 0.4307 | 0.0010 |

| 72 | 216 | 00:04:15 | 79.08% | 0.4297 | 0.0010 |

| 73 | 217 | 00:04:17 | 84.51% | 0.3789 | 0.0010 |

| 73 | 218 | 00:04:17 | 81.51% | 0.4362 | 0.0010 |

| 73 | 219 | 00:04:18 | 80.31% | 0.4152 | 0.0010 |

| 74 | 220 | 00:04:20 | 83.91% | 0.3807 | 0.0010 |

| 74 | 221 | 00:04:20 | 81.01% | 0.4260 | 0.0010 |

| 74 | 222 | 00:04:21 | 80.61% | 0.4178 | 0.0010 |

| 75 | 223 | 00:04:23 | 84.98% | 0.3784 | 0.0010 |

| 75 | 224 | 00:04:23 | 80.39% | 0.4361 | 0.0010 |

| 75 | 225 | 00:04:24 | 79.93% | 0.4108 | 0.0010 |

| 76 | 226 | 00:04:25 | 85.22% | 0.3881 | 0.0010 |

| 76 | 227 | 00:04:26 | 81.85% | 0.4229 | 0.0010 |

| 76 | 228 | 00:04:27 | 79.29% | 0.4279 | 0.0010 |

| 77 | 229 | 00:04:28 | 85.02% | 0.3676 | 0.0010 |

| 77 | 230 | 00:04:30 | 81.78% | 0.4339 | 0.0010 |

| 77 | 231 | 00:04:31 | 80.88% | 0.4020 | 0.0010 |

| 78 | 232 | 00:04:33 | 84.36% | 0.3710 | 0.0010 |

| 78 | 233 | 00:04:34 | 81.56% | 0.4181 | 0.0010 |

| 78 | 234 | 00:04:35 | 80.95% | 0.4096 | 0.0010 |

| 79 | 235 | 00:04:37 | 85.28% | 0.3676 | 0.0010 |

| 79 | 236 | 00:04:38 | 80.70% | 0.4328 | 0.0010 |

| 79 | 237 | 00:04:39 | 80.42% | 0.4021 | 0.0010 |

| 80 | 238 | 00:04:40 | 85.65% | 0.3760 | 0.0010 |

| 80 | 239 | 00:04:41 | 82.44% | 0.4156 | 0.0010 |

| 80 | 240 | 00:04:42 | 79.74% | 0.4231 | 0.0010 |

| 81 | 241 | 00:04:44 | 85.41% | 0.3561 | 0.0010 |

| 81 | 242 | 00:04:45 | 82.13% | 0.4259 | 0.0010 |

| 81 | 243 | 00:04:46 | 81.44% | 0.3921 | 0.0010 |

| 82 | 244 | 00:04:47 | 85.08% | 0.3573 | 0.0010 |

| 82 | 245 | 00:04:48 | 82.19% | 0.4094 | 0.0010 |

| 82 | 246 | 00:04:49 | 81.20% | 0.4039 | 0.0010 |

| 83 | 247 | 00:04:51 | 85.51% | 0.3547 | 0.0010 |

| 83 | 248 | 00:04:51 | 81.25% | 0.4305 | 0.0010 |

| 83 | 249 | 00:04:52 | 81.11% | 0.3909 | 0.0010 |

| 84 | 250 | 00:04:54 | 85.51% | 0.3703 | 0.0010 |

| 84 | 251 | 00:04:54 | 82.54% | 0.4098 | 0.0010 |

| 84 | 252 | 00:04:55 | 79.36% | 0.4274 | 0.0010 |

| 85 | 253 | 00:04:57 | 85.88% | 0.3527 | 0.0010 |

| 85 | 254 | 00:04:58 | 82.42% | 0.4138 | 0.0010 |

| 85 | 255 | 00:04:58 | 80.44% | 0.4001 | 0.0010 |

| 86 | 256 | 00:05:00 | 85.44% | 0.3462 | 0.0010 |

| 86 | 257 | 00:05:01 | 81.86% | 0.4164 | 0.0010 |

| 86 | 258 | 00:05:02 | 81.18% | 0.3892 | 0.0010 |

| 87 | 259 | 00:05:03 | 84.53% | 0.3577 | 0.0010 |

| 87 | 260 | 00:05:04 | 81.83% | 0.4158 | 0.0010 |

| 87 | 261 | 00:05:05 | 81.84% | 0.3791 | 0.0010 |

| 88 | 262 | 00:05:06 | 86.04% | 0.3427 | 0.0010 |

| 88 | 263 | 00:05:07 | 82.47% | 0.4041 | 0.0010 |

| 88 | 264 | 00:05:08 | 80.67% | 0.3993 | 0.0010 |

| 89 | 265 | 00:05:09 | 86.20% | 0.3396 | 0.0010 |

| 89 | 266 | 00:05:10 | 82.64% | 0.4103 | 0.0010 |

| 89 | 267 | 00:05:11 | 81.35% | 0.3895 | 0.0010 |

| 90 | 268 | 00:05:12 | 85.55% | 0.3375 | 0.0010 |

| 90 | 269 | 00:05:13 | 82.08% | 0.4166 | 0.0010 |

| 90 | 270 | 00:05:14 | 80.52% | 0.3962 | 0.0010 |

| 91 | 271 | 00:05:15 | 83.87% | 0.3643 | 0.0010 |

| 91 | 272 | 00:05:16 | 81.85% | 0.4197 | 0.0010 |

| 91 | 273 | 00:05:17 | 80.31% | 0.3977 | 0.0010 |

| 92 | 274 | 00:05:18 | 83.07% | 0.3756 | 0.0010 |

| 92 | 275 | 00:05:19 | 80.58% | 0.4314 | 0.0010 |

| 92 | 276 | 00:05:20 | 80.74% | 0.4002 | 0.0010 |

| 93 | 277 | 00:05:21 | 84.53% | 0.3484 | 0.0010 |

| 93 | 278 | 00:05:22 | 80.86% | 0.4295 | 0.0010 |

| 93 | 279 | 00:05:23 | 80.83% | 0.3985 | 0.0010 |

| 94 | 280 | 00:05:24 | 83.87% | 0.3600 | 0.0010 |

| 94 | 281 | 00:05:25 | 81.18% | 0.4311 | 0.0010 |

| 94 | 282 | 00:05:26 | 81.25% | 0.3899 | 0.0010 |

| 95 | 283 | 00:05:27 | 84.43% | 0.3747 | 0.0010 |

| 95 | 284 | 00:05:28 | 82.51% | 0.4139 | 0.0010 |

| 95 | 285 | 00:05:29 | 79.46% | 0.4221 | 0.0010 |

| 96 | 286 | 00:05:30 | 85.18% | 0.3456 | 0.0010 |

| 96 | 287 | 00:05:31 | 81.43% | 0.4362 | 0.0010 |

| 96 | 288 | 00:05:32 | 79.90% | 0.4256 | 0.0010 |

| 97 | 289 | 00:05:33 | 83.80% | 0.3726 | 0.0010 |

| 97 | 290 | 00:05:34 | 80.59% | 0.4527 | 0.0010 |

| 97 | 291 | 00:05:35 | 81.46% | 0.3864 | 0.0010 |

| 98 | 292 | 00:05:37 | 84.40% | 0.3562 | 0.0010 |

| 98 | 293 | 00:05:38 | 79.03% | 0.4471 | 0.0010 |

| 98 | 294 | 00:05:38 | 81.88% | 0.3806 | 0.0010 |

| 99 | 295 | 00:05:40 | 84.05% | 0.3586 | 0.0010 |

| 99 | 296 | 00:05:41 | 82.49% | 0.4050 | 0.0010 |

| 99 | 297 | 00:05:42 | 81.28% | 0.3977 | 0.0010 |

| 100 | 298 | 00:05:43 | 84.27% | 0.3523 | 0.0010 |

| 100 | 299 | 00:05:44 | 81.20% | 0.4274 | 0.0010 |

| 100 | 300 | 00:05:45 | 80.86% | 0.3957 | 0.0010 |

If you would like to check my code, I attached "Sajat.m"

I'm a completely beginner MATLAB user. Could someone help me?

Huge thanks in advance!

Answer 1

[Udit06]

Hi Peter,

In order to prevent the fluctuation in the results, you can try updating the learning rate adaptively while training the network. For achieving that you can refer to the "LearnRateSchedule", "LearnRateDropFactor" and "LearnRateDropPeriod" options of the "trainingOptions" that MATLAB provides.

https://www.mathworks.com/help/deeplearning/ref/trainingoptions.html

Other than that you can also use "imageDataAugmenter" to augment your dataset. This would be helpful since you only have 65 training images right now. Data augmentation can artificially expand the size of your training set by applying various transformations like rotation, scaling, flipping, and cropping, or even adding noise. This can help the network learn more robust features and improve its generalization capability. You can refer to the following MathWorks documentation to understand more about "imageDataAugmenter".

https://www.mathworks.com/help/deeplearning/ref/imagedataaugmenter.html

I hope this helps.