{
  "id": 479357,
  "title": "MelSpec GANs",
  "url": "/competitions/hms-harmful-brain-activity-classification/discussion/479357",
  "author_name": "Peter",
  "post_date": "2024-02-24T09:18:14.095000",
  "votes": 7,
  "comment_count": 0,
  "views": 0,
  "content": "<p>Hi, kagglers! </p>\n<p>I'm trying to build <strong>GANs</strong> Model</p>\n<pre><code>def build_generator(z_dim):\nmodel = Sequential()\nmodel.(Dense(256 * 32 * 64 , =z_dim))\nmodel.(Reshape((32,64,256))) \nmodel.(Conv2DTranspose(128, =3, =2, =))\nmodel.(BatchNormalization())\nmodel.(LeakyReLU(=0.01))\nmodel.(Conv2DTranspose(64, =3, =1, =))\nmodel.(BatchNormalization())\nmodel.(LeakyReLU(=0.01))\nmodel.(Conv2DTranspose(1, =3, =2, =))  \nmodel.(LeakyReLU(=0.01))\nreturn model\n\ndef build_discriminator(img_shape=(128,256,1)):\nmodel = Sequential()\nmodel.(\n     Conv2D(32,\n           =3,\n           =2,\n           =img_shape,\n           =)\n)\n\nmodel.(Activation())\n\nmodel.(\n   Conv2D(64,\n         =3,\n         =2,\n         =)\n)\n\nmodel.(Activation())\n\nmodel.(\n    Conv2D(128,\n         =3,\n         =2,\n         =))\n\nmodel.(Activation())\n\nmodel.(Flatten())\nmodel.(Dense(1, =))\n\nreturn model\n</code></pre>\n<p>But, unfortunately it is hard to training gan model haha😅<br>\nSo, I'm considering whether to continue with the GAN model or not<br>\nIs there any person who try to train gan models? Feel Free to tell me</p>",
  "messages": [
    {
      "id": 2666276,
      "postDate": "2024-02-24T09:18:14.097Z",
      "content": "<p>Hi, kagglers! </p>\n<p>I'm trying to build <strong>GANs</strong> Model</p>\n<pre><code>def build_generator(z_dim):\nmodel = Sequential()\nmodel.(Dense(256 * 32 * 64 , =z_dim))\nmodel.(Reshape((32,64,256))) \nmodel.(Conv2DTranspose(128, =3, =2, =))\nmodel.(BatchNormalization())\nmodel.(LeakyReLU(=0.01))\nmodel.(Conv2DTranspose(64, =3, =1, =))\nmodel.(BatchNormalization())\nmodel.(LeakyReLU(=0.01))\nmodel.(Conv2DTranspose(1, =3, =2, =))  \nmodel.(LeakyReLU(=0.01))\nreturn model\n\ndef build_discriminator(img_shape=(128,256,1)):\nmodel = Sequential()\nmodel.(\n     Conv2D(32,\n           =3,\n           =2,\n           =img_shape,\n           =)\n)\n\nmodel.(Activation())\n\nmodel.(\n   Conv2D(64,\n         =3,\n         =2,\n         =)\n)\n\nmodel.(Activation())\n\nmodel.(\n    Conv2D(128,\n         =3,\n         =2,\n         =))\n\nmodel.(Activation())\n\nmodel.(Flatten())\nmodel.(Dense(1, =))\n\nreturn model\n</code></pre>\n<p>But, unfortunately it is hard to training gan model haha😅<br>\nSo, I'm considering whether to continue with the GAN model or not<br>\nIs there any person who try to train gan models? Feel Free to tell me</p>",
      "rawMarkdown": "Hi, kagglers! \n\nI'm trying to build **GANs** Model\n\n    def build_generator(z_dim):\n    model = Sequential()\n    model.add(Dense(256 * 32 * 64 , input_dim=z_dim))\n    model.add(Reshape((32,64,256))) \n    model.add(Conv2DTranspose(128, kernel_size=3, strides=2, padding='same'))\n    model.add(BatchNormalization())\n    model.add(LeakyReLU(alpha=0.01))\n    model.add(Conv2DTranspose(64, kernel_size=3, strides=1, padding='same'))\n    model.add(BatchNormalization())\n    model.add(LeakyReLU(alpha=0.01))\n    model.add(Conv2DTranspose(1, kernel_size=3, strides=2, padding='same'))  \n    model.add(LeakyReLU(alpha=0.01))\n    return model\n\n    def build_discriminator(img_shape=(128,256,1)):\n    model = Sequential()\n    model.add(\n         Conv2D(32,\n               kernel_size=3,\n               strides=2,\n               input_shape=img_shape,\n               padding='same')\n    )\n    \n    model.add(Activation('tanh'))\n    \n    model.add(\n       Conv2D(64,\n             kernel_size=3,\n             strides=2,\n             padding='same')\n    )\n    \n    model.add(Activation('tanh'))\n    \n    model.add(\n        Conv2D(128,\n             kernel_size=3,\n             strides=2,\n             padding='same'))\n    \n    model.add(Activation('tanh'))\n    \n    model.add(Flatten())\n    model.add(Dense(1, activation='sigmoid'))\n    \n    return model\n\n\nBut, unfortunately it is hard to training gan model haha😅\nSo, I'm considering whether to continue with the GAN model or not\nIs there any person who try to train gan models? Feel Free to tell me",
      "votes": 7
    }
  ],
  "comments": [],
  "raw_markdown_by_id": {
    "2666276": "Hi, kagglers! \n\nI'm trying to build **GANs** Model\n\n    def build_generator(z_dim):\n    model = Sequential()\n    model.add(Dense(256 * 32 * 64 , input_dim=z_dim))\n    model.add(Reshape((32,64,256))) \n    model.add(Conv2DTranspose(128, kernel_size=3, strides=2, padding='same'))\n    model.add(BatchNormalization())\n    model.add(LeakyReLU(alpha=0.01))\n    model.add(Conv2DTranspose(64, kernel_size=3, strides=1, padding='same'))\n    model.add(BatchNormalization())\n    model.add(LeakyReLU(alpha=0.01))\n    model.add(Conv2DTranspose(1, kernel_size=3, strides=2, padding='same'))  \n    model.add(LeakyReLU(alpha=0.01))\n    return model\n\n    def build_discriminator(img_shape=(128,256,1)):\n    model = Sequential()\n    model.add(\n         Conv2D(32,\n               kernel_size=3,\n               strides=2,\n               input_shape=img_shape,\n               padding='same')\n    )\n    \n    model.add(Activation('tanh'))\n    \n    model.add(\n       Conv2D(64,\n             kernel_size=3,\n             strides=2,\n             padding='same')\n    )\n    \n    model.add(Activation('tanh'))\n    \n    model.add(\n        Conv2D(128,\n             kernel_size=3,\n             strides=2,\n             padding='same'))\n    \n    model.add(Activation('tanh'))\n    \n    model.add(Flatten())\n    model.add(Dense(1, activation='sigmoid'))\n    \n    return model\n\n\nBut, unfortunately it is hard to training gan model haha😅\nSo, I'm considering whether to continue with the GAN model or not\nIs there any person who try to train gan models? Feel Free to tell me"
  }
}