{
  "id": 248245,
  "title": "New Personal High Scores",
  "url": "/competitions/dogs-vs-cats/discussion/248245",
  "author_name": "djwillmsfl",
  "post_date": "2021-06-23T00:03:27.103000",
  "votes": 0,
  "comment_count": 0,
  "views": 0,
  "content": "<p>Accuracy = 99.979%  on 24198 images<br>\nErrors = 5 images of 24198 images<br>\nTrained on 13,000 images of 24198 images<br>\nPredicted on all 24198 images<br>\nBatch size = 30<br>\nInput image sizing = 150,150<br>\nClipping in Optimizer<br>\n50% drop out on most dense layers<br>\nSELU Activation - NOT RELU<br>\nSGD Optimizer<br>\nEpochs = 237 <br>\nTraining Run time = 1.5 days<br>\nGpu = Nvidia RTX 2070 super<br>\nTensorflow-gpu 2.5<br>\nKeras-gpu</p>\n<p>No scaling<br>\nNo Zooming<br>\nNo Rotation<br>\nNo Shear<br>\nNo Horizontal Flipping<br>\nNo Vertical Flipping<br>\n(However, even with no image alterations its trained on 13,000 of 24198 images and predicts on all 24198 with only 5 wrong classifications)</p>\n<p>Total params: 13,694,234<br>\nTrainable params: 13,694,234<br>\nNon-trainable params: 0</p>\n<p>model.add(Conv2D(256, (5, 5), strides=(2,2), padding=\"valid\",input_shape=input_shapes, activation='selu',kernel_regularizer=l2(0.0005)))<br>\n    model.add(Conv2D(180, (5, 5), strides=(1,1), padding=\"same\",input_shape=input_shapes, activation='selu',kernel_regularizer=l2(0.0005)))<br>\n    model.add(MaxPooling2D(pool_size=(2, 2)))</p>\n<pre><code>model.add(Conv2D(384, (4, 4), strides=(1,1), padding=\"same\",input_shape=input_shapes, activation='selu',kernel_regularizer=l2(0.0005)))\nmodel.add(MaxPooling2D(pool_size=(2, 2)))\n\nmodel.add(Conv2D(256, (2, 2), strides=(1,1), padding=\"same\",input_shape=input_shapes, activation='selu',kernel_regularizer=l2(0.0005)))\nmodel.add(MaxPooling2D(pool_size=(2, 2)))\n\nmodel.add(Flatten())\nmodel.add(Dense(512,kernel_initializer='random_uniform', activation='selu',kernel_regularizer=l2(0.0005)))\nmodel.add(Dense(512,kernel_initializer='random_uniform', activation='selu',kernel_regularizer=l2(0.0005)))\nmodel.add(Dropout(0.5))\nmodel.add(Dense(180,kernel_initializer='random_uniform', activation='selu',kernel_regularizer=l2(0.0005)))\nmodel.add(Dropout(0.5))\nmodel.add(Dense(256,kernel_initializer='random_uniform', activation='selu',kernel_regularizer=l2(0.0005)))\nmodel.add(Dropout(0.5))\nmodel.add(Dense(16,kernel_initializer='random_uniform', activation='selu',kernel_regularizer=l2(0.0005)))\nmodel.add(Dense(classes,activation='sigmoid'))\n</code></pre>\n<p>opt = SGD(learning_rate=0.015, decay=1e-08, momentum=0.3, nesterov=True, clipvalue=0.8)</p>\n<p>model.compile(loss=\"categorical_crossentropy\", optimizer=opt, metrics=['accuracy','categorical_accuracy','categorical_crossentropy'])</p>",
  "messages": [
    {
      "id": 1361593,
      "postDate": "2021-06-23T00:03:27.103Z",
      "content": "<p>Accuracy = 99.979%  on 24198 images<br>\nErrors = 5 images of 24198 images<br>\nTrained on 13,000 images of 24198 images<br>\nPredicted on all 24198 images<br>\nBatch size = 30<br>\nInput image sizing = 150,150<br>\nClipping in Optimizer<br>\n50% drop out on most dense layers<br>\nSELU Activation - NOT RELU<br>\nSGD Optimizer<br>\nEpochs = 237 <br>\nTraining Run time = 1.5 days<br>\nGpu = Nvidia RTX 2070 super<br>\nTensorflow-gpu 2.5<br>\nKeras-gpu</p>\n<p>No scaling<br>\nNo Zooming<br>\nNo Rotation<br>\nNo Shear<br>\nNo Horizontal Flipping<br>\nNo Vertical Flipping<br>\n(However, even with no image alterations its trained on 13,000 of 24198 images and predicts on all 24198 with only 5 wrong classifications)</p>\n<p>Total params: 13,694,234<br>\nTrainable params: 13,694,234<br>\nNon-trainable params: 0</p>\n<p>model.add(Conv2D(256, (5, 5), strides=(2,2), padding=\"valid\",input_shape=input_shapes, activation='selu',kernel_regularizer=l2(0.0005)))<br>\n    model.add(Conv2D(180, (5, 5), strides=(1,1), padding=\"same\",input_shape=input_shapes, activation='selu',kernel_regularizer=l2(0.0005)))<br>\n    model.add(MaxPooling2D(pool_size=(2, 2)))</p>\n<pre><code>model.add(Conv2D(384, (4, 4), strides=(1,1), padding=\"same\",input_shape=input_shapes, activation='selu',kernel_regularizer=l2(0.0005)))\nmodel.add(MaxPooling2D(pool_size=(2, 2)))\n\nmodel.add(Conv2D(256, (2, 2), strides=(1,1), padding=\"same\",input_shape=input_shapes, activation='selu',kernel_regularizer=l2(0.0005)))\nmodel.add(MaxPooling2D(pool_size=(2, 2)))\n\nmodel.add(Flatten())\nmodel.add(Dense(512,kernel_initializer='random_uniform', activation='selu',kernel_regularizer=l2(0.0005)))\nmodel.add(Dense(512,kernel_initializer='random_uniform', activation='selu',kernel_regularizer=l2(0.0005)))\nmodel.add(Dropout(0.5))\nmodel.add(Dense(180,kernel_initializer='random_uniform', activation='selu',kernel_regularizer=l2(0.0005)))\nmodel.add(Dropout(0.5))\nmodel.add(Dense(256,kernel_initializer='random_uniform', activation='selu',kernel_regularizer=l2(0.0005)))\nmodel.add(Dropout(0.5))\nmodel.add(Dense(16,kernel_initializer='random_uniform', activation='selu',kernel_regularizer=l2(0.0005)))\nmodel.add(Dense(classes,activation='sigmoid'))\n</code></pre>\n<p>opt = SGD(learning_rate=0.015, decay=1e-08, momentum=0.3, nesterov=True, clipvalue=0.8)</p>\n<p>model.compile(loss=\"categorical_crossentropy\", optimizer=opt, metrics=['accuracy','categorical_accuracy','categorical_crossentropy'])</p>",
      "rawMarkdown": "Accuracy = 99.979%  on 24198 images\nErrors = 5 images of 24198 images\nTrained on 13,000 images of 24198 images\nPredicted on all 24198 images\nBatch size = 30\nInput image sizing = 150,150\nClipping in Optimizer\n50% drop out on most dense layers\nSELU Activation - NOT RELU\nSGD Optimizer\nEpochs = 237 \nTraining Run time = 1.5 days\nGpu = Nvidia RTX 2070 super\nTensorflow-gpu 2.5\nKeras-gpu\n\nNo scaling\nNo Zooming\nNo Rotation\nNo Shear\nNo Horizontal Flipping\nNo Vertical Flipping\n(However, even with no image alterations its trained on 13,000 of 24198 images and predicts on all 24198 with only 5 wrong classifications)\n\nTotal params: 13,694,234\nTrainable params: 13,694,234\nNon-trainable params: 0\n\n  model.add(Conv2D(256, (5, 5), strides=(2,2), padding=\"valid\",input_shape=input_shapes, activation='selu',kernel_regularizer=l2(0.0005)))\n    model.add(Conv2D(180, (5, 5), strides=(1,1), padding=\"same\",input_shape=input_shapes, activation='selu',kernel_regularizer=l2(0.0005)))\n    model.add(MaxPooling2D(pool_size=(2, 2)))\n     \n    model.add(Conv2D(384, (4, 4), strides=(1,1), padding=\"same\",input_shape=input_shapes, activation='selu',kernel_regularizer=l2(0.0005)))\n    model.add(MaxPooling2D(pool_size=(2, 2)))\n\n    model.add(Conv2D(256, (2, 2), strides=(1,1), padding=\"same\",input_shape=input_shapes, activation='selu',kernel_regularizer=l2(0.0005)))\n    model.add(MaxPooling2D(pool_size=(2, 2)))\n    \n    model.add(Flatten())\n    model.add(Dense(512,kernel_initializer='random_uniform', activation='selu',kernel_regularizer=l2(0.0005)))\n    model.add(Dense(512,kernel_initializer='random_uniform', activation='selu',kernel_regularizer=l2(0.0005)))\n    model.add(Dropout(0.5))\n    model.add(Dense(180,kernel_initializer='random_uniform', activation='selu',kernel_regularizer=l2(0.0005)))\n    model.add(Dropout(0.5))\n    model.add(Dense(256,kernel_initializer='random_uniform', activation='selu',kernel_regularizer=l2(0.0005)))\n    model.add(Dropout(0.5))\n    model.add(Dense(16,kernel_initializer='random_uniform', activation='selu',kernel_regularizer=l2(0.0005)))\n    model.add(Dense(classes,activation='sigmoid'))\n\nopt = SGD(learning_rate=0.015, decay=1e-08, momentum=0.3, nesterov=True, clipvalue=0.8)\n\nmodel.compile(loss=\"categorical_crossentropy\", optimizer=opt, metrics=['accuracy','categorical_accuracy','categorical_crossentropy'])\n\n"
    }
  ],
  "comments": [],
  "raw_markdown_by_id": {
    "1361593": "Accuracy = 99.979%  on 24198 images\nErrors = 5 images of 24198 images\nTrained on 13,000 images of 24198 images\nPredicted on all 24198 images\nBatch size = 30\nInput image sizing = 150,150\nClipping in Optimizer\n50% drop out on most dense layers\nSELU Activation - NOT RELU\nSGD Optimizer\nEpochs = 237 \nTraining Run time = 1.5 days\nGpu = Nvidia RTX 2070 super\nTensorflow-gpu 2.5\nKeras-gpu\n\nNo scaling\nNo Zooming\nNo Rotation\nNo Shear\nNo Horizontal Flipping\nNo Vertical Flipping\n(However, even with no image alterations its trained on 13,000 of 24198 images and predicts on all 24198 with only 5 wrong classifications)\n\nTotal params: 13,694,234\nTrainable params: 13,694,234\nNon-trainable params: 0\n\n  model.add(Conv2D(256, (5, 5), strides=(2,2), padding=\"valid\",input_shape=input_shapes, activation='selu',kernel_regularizer=l2(0.0005)))\n    model.add(Conv2D(180, (5, 5), strides=(1,1), padding=\"same\",input_shape=input_shapes, activation='selu',kernel_regularizer=l2(0.0005)))\n    model.add(MaxPooling2D(pool_size=(2, 2)))\n     \n    model.add(Conv2D(384, (4, 4), strides=(1,1), padding=\"same\",input_shape=input_shapes, activation='selu',kernel_regularizer=l2(0.0005)))\n    model.add(MaxPooling2D(pool_size=(2, 2)))\n\n    model.add(Conv2D(256, (2, 2), strides=(1,1), padding=\"same\",input_shape=input_shapes, activation='selu',kernel_regularizer=l2(0.0005)))\n    model.add(MaxPooling2D(pool_size=(2, 2)))\n    \n    model.add(Flatten())\n    model.add(Dense(512,kernel_initializer='random_uniform', activation='selu',kernel_regularizer=l2(0.0005)))\n    model.add(Dense(512,kernel_initializer='random_uniform', activation='selu',kernel_regularizer=l2(0.0005)))\n    model.add(Dropout(0.5))\n    model.add(Dense(180,kernel_initializer='random_uniform', activation='selu',kernel_regularizer=l2(0.0005)))\n    model.add(Dropout(0.5))\n    model.add(Dense(256,kernel_initializer='random_uniform', activation='selu',kernel_regularizer=l2(0.0005)))\n    model.add(Dropout(0.5))\n    model.add(Dense(16,kernel_initializer='random_uniform', activation='selu',kernel_regularizer=l2(0.0005)))\n    model.add(Dense(classes,activation='sigmoid'))\n\nopt = SGD(learning_rate=0.015, decay=1e-08, momentum=0.3, nesterov=True, clipvalue=0.8)\n\nmodel.compile(loss=\"categorical_crossentropy\", optimizer=opt, metrics=['accuracy','categorical_accuracy','categorical_crossentropy'])\n\n"
  }
}