ex08_다중분류 모델 만들기(3종류 동물).ipynb

{"nbformat":4,"nbformat_minor":0,"metadata":{"colab":{"provenance":[],"collapsed_sections":[],"mount_file_id":"14WebiDMrI6D-03bY-HJYggHanDeA9fDY","authorship_tag":"ABX9TyPF89VCYKAu5TDFLi4phQNJ"},"kernelspec":{"name":"python3","display_name":"Python 3"},"language_info":{"name":"python"},"accelerator":"GPU"},"cells":[{"cell_type":"markdown","source":["## 목표\n","- 만들어진 이미지 데이터 셋 (고슴도치, 햄스터, 푸들 이미지)으로 다중 분류 모델 생성하기\n","\n","### 생성한 NPZ 파일 로드"],"metadata":{"id":"sK_zE7Z0-DCV"}},{"cell_type":"code","source":["import pandas as pd\n","import numpy as np\n","import matplotlib.pyplot as plt\n","from tensorflow.keras import Sequential\n","from tensorflow.keras.layers import Dense, Flatten"],"metadata":{"id":"9T0HvibE-h5X"},"execution_count":null,"outputs":[]},{"cell_type":"code","source":["data = np.load('/content/drive/MyDrive/Colab Notebooks/2022GJAI_DL/GJAI_DL/data/animal.npz')"],"metadata":{"id":"gvlzDobe-5-v"},"execution_count":null,"outputs":[]},{"cell_type":"code","source":["X_train = data['X_train']\n","X_test = data['X_test']\n","y_train = data['y_train']\n","y_test = data['y_test']\n","\n","X_train.shape, X_test.shape, y_train.shape, y_test.shape"],"metadata":{"colab":{"base_uri":"https://localhost:8080/"},"id":"o_G-Vb8L_jIJ","executionInfo":{"status":"ok","timestamp":1665558819557,"user_tz":-540,"elapsed":880,"user":{"displayName":"민윤기","userId":"13288389921392377597"}},"outputId":"9d933552-47e0-453e-84a9-2a8423de5e84"},"execution_count":null,"outputs":[{"output_type":"execute_result","data":{"text/plain":["((960, 224, 224, 3), (240, 224, 224, 3), (960,), (240,))"]},"metadata":{},"execution_count":7}]},{"cell_type":"markdown","source":["### MLP(다층퍼셉트론) 모델링"],"metadata":{"id":"qxSTwTnLCm3_"}},{"cell_type":"code","source":["model = Sequential()\n","\n","model.add(Flatten(input_shape=(224,224,3)))\n","model.add(Dense(64, activation='relu'))\n","model.add(Dense(256, activation='relu'))\n","model.add(Dense(128, activation='relu'))\n","model.add(Dense(32, activation='sigmoid'))\n","model.add(Dense(3, activation='softmax'))"],"metadata":{"id":"wRuiXq6QCvTN"},"execution_count":null,"outputs":[]},{"cell_type":"code","source":["model.compile(loss=\"sparse_categorical_crossentropy\",\n","              optimizer=\"Adam\",\n","              metrics=['acc']\n","              )"],"metadata":{"id":"6x-X6rVYDvVx"},"execution_count":null,"outputs":[]},{"cell_type":"code","source":["h = model.fit(X_train, y_train, validation_split=0.2, epochs=50, batch_size=128)"],"metadata":{"colab":{"base_uri":"https://localhost:8080/"},"id":"F1hu9VhDD1M9","executionInfo":{"status":"ok","timestamp":1665559305766,"user_tz":-540,"elapsed":10418,"user":{"displayName":"민윤기","userId":"13288389921392377597"}},"outputId":"91216c94-cb21-4f37-f7a0-5fd1d70ea3b5"},"execution_count":null,"outputs":[{"output_type":"stream","name":"stdout","text":["Epoch 1/50\n","6/6 [==============================] - 1s 57ms/step - loss: 1.3755 - acc: 0.3138 - val_loss: 1.2393 - val_acc: 0.3698\n","Epoch 2/50\n","6/6 [==============================] - 0s 21ms/step - loss: 1.3049 - acc: 0.3190 - val_loss: 1.3064 - val_acc: 0.3698\n","Epoch 3/50\n","6/6 [==============================] - 0s 21ms/step - loss: 1.2676 - acc: 0.3294 - val_loss: 1.1620 - val_acc: 0.2969\n","Epoch 4/50\n","6/6 [==============================] - 0s 22ms/step - loss: 1.1388 - acc: 0.3424 - val_loss: 1.1469 - val_acc: 0.2969\n","Epoch 5/50\n","6/6 [==============================] - 0s 23ms/step - loss: 1.1292 - acc: 0.3281 - val_loss: 1.0988 - val_acc: 0.3333\n","Epoch 6/50\n","6/6 [==============================] - 0s 21ms/step - loss: 1.0998 - acc: 0.3385 - val_loss: 1.1004 - val_acc: 0.3281\n","Epoch 7/50\n","6/6 [==============================] - 0s 29ms/step - loss: 1.1028 - acc: 0.3320 - val_loss: 1.1001 - val_acc: 0.3333\n","Epoch 8/50\n","6/6 [==============================] - 0s 22ms/step - loss: 1.1012 - acc: 0.3385 - val_loss: 1.1007 - val_acc: 0.3333\n","Epoch 9/50\n","6/6 [==============================] - 0s 22ms/step - loss: 1.1004 - acc: 0.3385 - val_loss: 1.1007 - val_acc: 0.3333\n","Epoch 10/50\n","6/6 [==============================] - 0s 22ms/step - loss: 1.0995 - acc: 0.3385 - val_loss: 1.1009 - val_acc: 0.3333\n","Epoch 11/50\n","6/6 [==============================] - 0s 23ms/step - loss: 1.0989 - acc: 0.3385 - val_loss: 1.1010 - val_acc: 0.3333\n","Epoch 12/50\n","6/6 [==============================] - 0s 22ms/step - loss: 1.0986 - acc: 0.3385 - val_loss: 1.1019 - val_acc: 0.3333\n","Epoch 13/50\n","6/6 [==============================] - 0s 25ms/step - loss: 1.0984 - acc: 0.3385 - val_loss: 1.1018 - val_acc: 0.3333\n","Epoch 14/50\n","6/6 [==============================] - 0s 22ms/step - loss: 1.0982 - acc: 0.3529 - val_loss: 1.1016 - val_acc: 0.2969\n","Epoch 15/50\n","6/6 [==============================] - 0s 23ms/step - loss: 1.0982 - acc: 0.3424 - val_loss: 1.1020 - val_acc: 0.2969\n","Epoch 16/50\n","6/6 [==============================] - 0s 22ms/step - loss: 1.0982 - acc: 0.3424 - val_loss: 1.1022 - val_acc: 0.2969\n","Epoch 17/50\n","6/6 [==============================] - 0s 22ms/step - loss: 1.0983 - acc: 0.3424 - val_loss: 1.1021 - val_acc: 0.2969\n","Epoch 18/50\n","6/6 [==============================] - 0s 22ms/step - loss: 1.0983 - acc: 0.3424 - val_loss: 1.1025 - val_acc: 0.2969\n","Epoch 19/50\n","6/6 [==============================] - 0s 23ms/step - loss: 1.0982 - acc: 0.3424 - val_loss: 1.1021 - val_acc: 0.2969\n","Epoch 20/50\n","6/6 [==============================] - 0s 24ms/step - loss: 1.0982 - acc: 0.3424 - val_loss: 1.1019 - val_acc: 0.2969\n","Epoch 21/50\n","6/6 [==============================] - 0s 22ms/step - loss: 1.0982 - acc: 0.3424 - val_loss: 1.1016 - val_acc: 0.2969\n","Epoch 22/50\n","6/6 [==============================] - 0s 23ms/step - loss: 1.0982 - acc: 0.3424 - val_loss: 1.1016 - val_acc: 0.2969\n","Epoch 23/50\n","6/6 [==============================] - 0s 25ms/step - loss: 1.0982 - acc: 0.3424 - val_loss: 1.1017 - val_acc: 0.2969\n","Epoch 24/50\n","6/6 [==============================] - 0s 22ms/step - loss: 1.0982 - acc: 0.3424 - val_loss: 1.1017 - val_acc: 0.2969\n","Epoch 25/50\n","6/6 [==============================] - 0s 22ms/step - loss: 1.0982 - acc: 0.3424 - val_loss: 1.1015 - val_acc: 0.2969\n","Epoch 26/50\n","6/6 [==============================] - 0s 23ms/step - loss: 1.0982 - acc: 0.3424 - val_loss: 1.1016 - val_acc: 0.2969\n","Epoch 27/50\n","6/6 [==============================] - 0s 21ms/step - loss: 1.0983 - acc: 0.3424 - val_loss: 1.1020 - val_acc: 0.2969\n","Epoch 28/50\n","6/6 [==============================] - 0s 22ms/step - loss: 1.0982 - acc: 0.3424 - val_loss: 1.1016 - val_acc: 0.2969\n","Epoch 29/50\n","6/6 [==============================] - 0s 22ms/step - loss: 1.0982 - acc: 0.3424 - val_loss: 1.1015 - val_acc: 0.2969\n","Epoch 30/50\n","6/6 [==============================] - 0s 22ms/step - loss: 1.0982 - acc: 0.3424 - val_loss: 1.1014 - val_acc: 0.2969\n","Epoch 31/50\n","6/6 [==============================] - 0s 22ms/step - loss: 1.0982 - acc: 0.3424 - val_loss: 1.1016 - val_acc: 0.2969\n","Epoch 32/50\n","6/6 [==============================] - 0s 22ms/step - loss: 1.0983 - acc: 0.3424 - val_loss: 1.1013 - val_acc: 0.2969\n","Epoch 33/50\n","6/6 [==============================] - 0s 22ms/step - loss: 1.0983 - acc: 0.3424 - val_loss: 1.1013 - val_acc: 0.2969\n","Epoch 34/50\n","6/6 [==============================] - 0s 22ms/step - loss: 1.0982 - acc: 0.3424 - val_loss: 1.1016 - val_acc: 0.2969\n","Epoch 35/50\n","6/6 [==============================] - 0s 25ms/step - loss: 1.0982 - acc: 0.3424 - val_loss: 1.1020 - val_acc: 0.2969\n","Epoch 36/50\n","6/6 [==============================] - 0s 23ms/step - loss: 1.0986 - acc: 0.3424 - val_loss: 1.1019 - val_acc: 0.2969\n","Epoch 37/50\n","6/6 [==============================] - 0s 22ms/step - loss: 1.0983 - acc: 0.3424 - val_loss: 1.1022 - val_acc: 0.2969\n","Epoch 38/50\n","6/6 [==============================] - 0s 22ms/step - loss: 1.0982 - acc: 0.3424 - val_loss: 1.1019 - val_acc: 0.2969\n","Epoch 39/50\n","6/6 [==============================] - 0s 22ms/step - loss: 1.0982 - acc: 0.3216 - val_loss: 1.1018 - val_acc: 0.3333\n","Epoch 40/50\n","6/6 [==============================] - 0s 21ms/step - loss: 1.0983 - acc: 0.3086 - val_loss: 1.1014 - val_acc: 0.3333\n","Epoch 41/50\n","6/6 [==============================] - 0s 22ms/step - loss: 1.0983 - acc: 0.3281 - val_loss: 1.1020 - val_acc: 0.2969\n","Epoch 42/50\n","6/6 [==============================] - 0s 24ms/step - loss: 1.0982 - acc: 0.3424 - val_loss: 1.1020 - val_acc: 0.2969\n","Epoch 43/50\n","6/6 [==============================] - 0s 23ms/step - loss: 1.0983 - acc: 0.3138 - val_loss: 1.1016 - val_acc: 0.2969\n","Epoch 44/50\n","6/6 [==============================] - 0s 22ms/step - loss: 1.0983 - acc: 0.3424 - val_loss: 1.1013 - val_acc: 0.2969\n","Epoch 45/50\n","6/6 [==============================] - 0s 20ms/step - loss: 1.0982 - acc: 0.3424 - val_loss: 1.1012 - val_acc: 0.2969\n","Epoch 46/50\n","6/6 [==============================] - 0s 22ms/step - loss: 1.0982 - acc: 0.3424 - val_loss: 1.1015 - val_acc: 0.2969\n","Epoch 47/50\n","6/6 [==============================] - 0s 22ms/step - loss: 1.0982 - acc: 0.3424 - val_loss: 1.1018 - val_acc: 0.2969\n","Epoch 48/50\n","6/6 [==============================] - 0s 22ms/step - loss: 1.0983 - acc: 0.3424 - val_loss: 1.1016 - val_acc: 0.2969\n","Epoch 49/50\n","6/6 [==============================] - 0s 21ms/step - loss: 1.0982 - acc: 0.3424 - val_loss: 1.1017 - val_acc: 0.2969\n","Epoch 50/50\n","6/6 [==============================] - 0s 23ms/step - loss: 1.0982 - acc: 0.3424 - val_loss: 1.1017 - val_acc: 0.2969\n"]}]},{"cell_type":"code","source":["plt.figure(figsize=(15, 5))\n","\n","plt.plot(h.history['acc'], label='acc')\n","plt.plot(h.history['val_acc'], label='val_acc')\n","plt.legend()\n","plt.show()"],"metadata":{"colab":{"base_uri":"https://localhost:8080/","height":320},"id":"9YT1rjXhD-td","executionInfo":{"status":"ok","timestamp":1665559405375,"user_tz":-540,"elapsed":1071,"user":{"displayName":"민윤기","userId":"13288389921392377597"}},"outputId":"5d0b5a3a-d5ed-4cad-c954-1d06a4d97181"},"execution_count":null,"outputs":[{"output_type":"display_data","data":{"text/plain":["<Figure size 1080x360 with 1 Axes>"],"image/png":"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\n"},"metadata":{"needs_background":"light"}}]},{"cell_type":"markdown","source":["### MLP 이미지 분석의 한계\n","- 결과를 보면 MLP로는 해당 데이터 셋으로 분류하기가 어렵다고 확인할 수 있음\n","- MLP 신경망은 이미지의 위치에 민감하게 동작하며 위치에 종속적인 결과를 얻게 됨\n","- 모든 픽셀을 연산하기 때문이다\n","- MLP로 이러한 데이터를 인식하려면 숫자의 크기와 위치를 비슷하게 맞춰줘야 함\n","- 이러한 데이터는 CNN으로 해결"],"metadata":{"id":"cCK1OM4WE5je"}},{"cell_type":"code","source":["cd /content/drive/MyDrive/Colab Notebooks/2022GJAI_DL/GJAI_DL"],"metadata":{"colab":{"base_uri":"https://localhost:8080/"},"id":"F5c6Rq3mkYVH","executionInfo":{"status":"ok","timestamp":1665567676925,"user_tz":-540,"elapsed":6,"user":{"displayName":"민윤기","userId":"13288389921392377597"}},"outputId":"0580e59b-a162-45fa-bf40-353f29542762"},"execution_count":9,"outputs":[{"output_type":"stream","name":"stdout","text":["/content/drive/MyDrive/Colab Notebooks/2022GJAI_DL/GJAI_DL\n"]}]},{"cell_type":"code","source":["ls -al"],"metadata":{"colab":{"base_uri":"https://localhost:8080/"},"id":"R3eulsBUkdWq","executionInfo":{"status":"ok","timestamp":1665567683019,"user_tz":-540,"elapsed":1139,"user":{"displayName":"민윤기","userId":"13288389921392377597"}},"outputId":"68223ec6-859e-4b32-e908-8c7a80059664"},"execution_count":10,"outputs":[{"output_type":"stream","name":"stdout","text":["total 447\n","drwx------ 5 root root   4096 Oct  6 06:52  \u001b[0m\u001b[01;34mdata\u001b[0m/\n","-rw------- 1 root root  43022 Oct  7 03:24  ex01.ipynb\n","-rw------- 1 root root   3346 Oct  7 05:40  ex02_퍼셉트론_XOR구현.ipynb\n","-rw------- 1 root root  47086 Oct 11 02:48  ex03_이진분류.ipynb\n","-rw------- 1 root root  47735 Oct 12 00:29 'ex04_이미지_다중분류(손글씨데이터).ipynb'\n","-rw------- 1 root root  61806 Oct 11 07:42 'ex05_이미지_다중분류(패션데이터).ipynb'\n","-rw------- 1 root root  73878 Oct 12 05:53  ex06_얼굴이미지_이진분류.ipynb\n","-rw------- 1 root root 117935 Oct 12 05:48 'ex07_활성화함수, 경사하강법 비교.ipynb'\n","-rw------- 1 root root  39538 Oct 12 09:40 'ex08_다중분류 모델 만들기(3종류 동물).ipynb'\n","drwx------ 6 root root   4096 Oct 12 09:24  \u001b[01;34m.git\u001b[0m/\n","drwx------ 2 root root   4096 Oct 12 01:16  \u001b[01;34mimage\u001b[0m/\n","drwx------ 2 root root   4096 Oct 12 01:15  \u001b[01;34m.ipynb_checkpoints\u001b[0m/\n","drwx------ 2 root root   4096 Oct 12 03:32  \u001b[01;34mmodel\u001b[0m/\n"]}]},{"cell_type":"code","source":["!git config --global user.email \"zoobon63@gmail.com\"\n","!git config --global user.name \"minyoongi96\"\n","!git add .\n","!git commit -m \"1012\"\n","!git push origin master"],"metadata":{"colab":{"base_uri":"https://localhost:8080/"},"id":"XZDSB1ipkemu","executionInfo":{"status":"ok","timestamp":1665568075111,"user_tz":-540,"elapsed":8664,"user":{"displayName":"민윤기","userId":"13288389921392377597"}},"outputId":"4df376c9-ef5c-4d77-afed-643ee092d3d3"},"execution_count":16,"outputs":[{"output_type":"stream","name":"stdout","text":["[master 1775b10] 1012\n"," 1 file changed, 1 insertion(+), 1 deletion(-)\n","Counting objects: 13, done.\n","Delta compression using up to 2 threads.\n","Compressing objects: 100% (12/12), done.\n","Writing objects: 100% (13/13), 107.96 MiB | 18.99 MiB/s, done.\n","Total 13 (delta 6), reused 0 (delta 0)\n","remote: Resolving deltas: 100% (6/6), completed with 4 local objects.\u001b[K\n","remote: error: Trace: eaeb6a8951f27bc5b31c1af9672f14a9347330596037534844ac987e08d6c775\u001b[K\n","remote: error: See http://git.io/iEPt8g for more information.\u001b[K\n","remote: error: File data/animal.npz is 107.73 MB; this exceeds GitHub's file size limit of 100.00 MB\u001b[K\n","remote: error: GH001: Large files detected. You may want to try Git Large File Storage - https://git-lfs.github.com.\u001b[K\n","To https://github.com/minyoongi96/GJAI_DL.git\n"," ! [remote rejected] master -> master (pre-receive hook declined)\n","error: failed to push some refs to 'https://minyoongi96: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@github.com/minyoongi96/GJAI_DL.git'\n"]}]},{"cell_type":"code","source":["!git filter-branch -f --index-filter 'git rm --cached --ignore-unmatch /content/drive/MyDrive/Colab Notebooks/2022GJAI_DL/GJAI_DL/data/animal.npz'"],"metadata":{"colab":{"base_uri":"https://localhost:8080/"},"id":"QxotoGIfkk24","executionInfo":{"status":"ok","timestamp":1665568080972,"user_tz":-540,"elapsed":816,"user":{"displayName":"민윤기","userId":"13288389921392377597"}},"outputId":"a8964326-a48f-45a0-c23a-e88ae57d5be9"},"execution_count":17,"outputs":[{"output_type":"stream","name":"stdout","text":["\rRewrite a477f4bb732275f8d1e5a46239d440bb75bfa92f (1/6) (0 seconds passed, remaining 0 predicted)    fatal: /content/drive/MyDrive/Colab: '/content/drive/MyDrive/Colab' is outside repository\n","index filter failed: git rm --cached --ignore-unmatch /content/drive/MyDrive/Colab Notebooks/2022GJAI_DL/GJAI_DL/data/animal.npz\n"]}]},{"cell_type":"code","source":["!git push origin master"],"metadata":{"colab":{"base_uri":"https://localhost:8080/"},"id":"5JO7pxjYlu-f","executionInfo":{"status":"ok","timestamp":1665568094372,"user_tz":-540,"elapsed":7843,"user":{"displayName":"민윤기","userId":"13288389921392377597"}},"outputId":"731cd084-2f40-4aa6-84c7-860ff485f21f"},"execution_count":18,"outputs":[{"output_type":"stream","name":"stdout","text":["Counting objects: 13, done.\n","Delta compression using up to 2 threads.\n","Compressing objects: 100% (12/12), done.\n","Writing objects: 100% (13/13), 107.96 MiB | 19.28 MiB/s, done.\n","Total 13 (delta 6), reused 0 (delta 0)\n","remote: Resolving deltas: 100% (6/6), completed with 4 local objects.\u001b[K\n","remote: error: Trace: c6c31004100674fd6e41944188a8896130e879e991a931494e6224df8d07c791\u001b[K\n","remote: error: See http://git.io/iEPt8g for more information.\u001b[K\n","remote: error: File data/animal.npz is 107.73 MB; this exceeds GitHub's file size limit of 100.00 MB\u001b[K\n","remote: error: GH001: Large files detected. You may want to try Git Large File Storage - https://git-lfs.github.com.\u001b[K\n","To https://github.com/minyoongi96/GJAI_DL.git\n"," ! [remote rejected] master -> master (pre-receive hook declined)\n","error: failed to push some refs to 'https://minyoongi96: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@github.com/minyoongi96/GJAI_DL.git'\n"]}]},{"cell_type":"code","source":["pwd"],"metadata":{"id":"9jVqtZ2LlwNk"},"execution_count":null,"outputs":[]}]}