nbs/keras_raw-vgg16.ipynb

keras_raw-vgg16.ipynb

{
  "cells": [
    {
      "metadata": {},
      "cell_type": "markdown",
      "source": "## Start"
    },
    {
      "metadata": {
        "collapsed": true,
        "trusted": true
      },
      "cell_type": "code",
      "source": "%reload_ext autoreload\n%autoreload 2\n%matplotlib inline",
      "execution_count": 1,
      "outputs": []
    },
    {
      "metadata": {
        "trusted": true
      },
      "cell_type": "code",
      "source": "from imports import *\nfrom fast_gen import preprocess_scale, scale_and_center",
      "execution_count": 1,
      "outputs": [
        {
          "name": "stderr",
          "output_type": "stream",
          "text": "Using TensorFlow backend.\n"
        }
      ]
    },
    {
      "metadata": {
        "collapsed": true,
        "trusted": true
      },
      "cell_type": "code",
      "source": "imagenet_mean = np.array([103.939, 116.779, 123.68], dtype=np.float32).reshape((1,1,3))\ndef preprocess_imagenet(x): return x[..., ::-1] - imagenet_mean",
      "execution_count": 13,
      "outputs": []
    },
    {
      "metadata": {
        "collapsed": true,
        "trusted": true
      },
      "cell_type": "code",
      "source": "bs=64; sz=224; lr=2e-3",
      "execution_count": 3,
      "outputs": []
    },
    {
      "metadata": {
        "collapsed": true,
        "trusted": true
      },
      "cell_type": "code",
      "source": "path = \"/data/jhoward/fast/dogscats/\"",
      "execution_count": 4,
      "outputs": []
    },
    {
      "metadata": {
        "collapsed": true,
        "trusted": true
      },
      "cell_type": "code",
      "source": "mn=VGG16()\nconv_outp=mn.get_layer('predictions').input\noutp = Dense(2, activation='softmax')(conv_outp)\nm = Model(mn.input, outp)\nfor l in m.layers[:-1]: l.trainable=False\nm.compile(SGD(lr, momentum=0.9), 'categorical_crossentropy', metrics=['accuracy'])",
      "execution_count": 5,
      "outputs": []
    },
    {
      "metadata": {
        "scrolled": false,
        "trusted": true
      },
      "cell_type": "code",
      "source": "gen = image.ImageDataGenerator(preprocessing_function=preprocess_imagenet)\ntrn_batches = gen.flow_from_directory(f'{path}train', (sz,sz), batch_size=bs)\nfix_batches = gen.flow_from_directory(f'{path}train', (sz,sz), batch_size=bs, shuffle=False)\nval_batches = gen.flow_from_directory(f'{path}valid', (sz,sz), batch_size=bs, shuffle=False)\nnb_trn = math.ceil(trn_batches.n/bs)\nnb_val = math.ceil(val_batches.n/bs)",
      "execution_count": 9,
      "outputs": [
        {
          "name": "stdout",
          "output_type": "stream",
          "text": "Found 23000 images belonging to 2 classes.\nFound 23000 images belonging to 2 classes.\nFound 2000 images belonging to 2 classes.\n"
        }
      ]
    },
    {
      "metadata": {
        "scrolled": false,
        "trusted": true
      },
      "cell_type": "code",
      "source": "m.fit_generator(trn_batches, nb_trn, workers=1, epochs=3,\n                 validation_data=val_batches, validation_steps=nb_val)",
      "execution_count": 10,
      "outputs": [
        {
          "name": "stdout",
          "output_type": "stream",
          "text": "Epoch 1/3\n360/360 [==============================] - 136s - loss: 0.0698 - acc: 0.9772 - val_loss: 0.0515 - val_acc: 0.9805\nEpoch 2/3\n360/360 [==============================] - 132s - loss: 0.0400 - acc: 0.9859 - val_loss: 0.0441 - val_acc: 0.9830\nEpoch 3/3\n360/360 [==============================] - 132s - loss: 0.0294 - acc: 0.9891 - val_loss: 0.0425 - val_acc: 0.9850\n"
        },
        {
          "data": {
            "text/plain": "<keras.callbacks.History at 0x7facf4f38160>"
          },
          "execution_count": 10,
          "metadata": {},
          "output_type": "execute_result"
        }
      ]
    },
    {
      "metadata": {
        "collapsed": true,
        "trusted": true
      },
      "cell_type": "code",
      "source": "",
      "execution_count": null,
      "outputs": []
    },
    {
      "metadata": {
        "collapsed": true,
        "trusted": true
      },
      "cell_type": "code",
      "source": "mn=VGG16()\nconv_outp=mn.get_layer('predictions').input\nm = Model(mn.input, conv_outp)",
      "execution_count": 17,
      "outputs": []
    },
    {
      "metadata": {
        "trusted": true
      },
      "cell_type": "code",
      "source": "fix_batches.reset(); val_batches.reset()\ntrn_acts = m.predict_generator(generator=fix_batches, verbose=1, \n                                steps=nb_trn, workers=1)\nval_acts = m.predict_generator(generator=val_batches, verbose=1,\n                                steps=nb_val, workers=1)",
      "execution_count": 18,
      "outputs": [
        {
          "name": "stdout",
          "output_type": "stream",
          "text": "360/360 [==============================] - 121s   \n32/32 [==============================] - 10s    \n"
        }
      ]
    },
    {
      "metadata": {
        "trusted": true,
        "collapsed": true
      },
      "cell_type": "code",
      "source": "inp = Input(batch_shape=m.output_shape)\noutp = Dense(1, activation='sigmoid')(inp)\nfc = Model(inp, outp)\nfc.compile(SGD(lr, momentum=0.9), 'binary_crossentropy', metrics=['accuracy'])",
      "execution_count": 20,
      "outputs": []
    },
    {
      "metadata": {
        "trusted": true
      },
      "cell_type": "code",
      "source": "fc.fit(trn_acts, fix_batches.classes, bs, 3, validation_data=(val_acts, val_batches.classes))",
      "execution_count": 21,
      "outputs": [
        {
          "name": "stdout",
          "output_type": "stream",
          "text": "Train on 23000 samples, validate on 2000 samples\nEpoch 1/3\n23000/23000 [==============================] - 1s - loss: 0.0605 - acc: 0.9773 - val_loss: 0.0556 - val_acc: 0.9820\nEpoch 2/3\n23000/23000 [==============================] - 1s - loss: 0.0375 - acc: 0.9857 - val_loss: 0.0378 - val_acc: 0.9845\nEpoch 3/3\n23000/23000 [==============================] - 1s - loss: 0.0306 - acc: 0.9888 - val_loss: 0.0366 - val_acc: 0.9845\n"
        },
        {
          "data": {
            "text/plain": "<keras.callbacks.History at 0x7faceda16c50>"
          },
          "execution_count": 21,
          "metadata": {},
          "output_type": "execute_result"
        }
      ]
    },
    {
      "metadata": {
        "collapsed": true,
        "trusted": true
      },
      "cell_type": "code",
      "source": "",
      "execution_count": null,
      "outputs": []
    }
  ],
  "metadata": {
    "kernelspec": {
      "name": "conda-root-py",
      "display_name": "Python [conda root]",
      "language": "python"
    },
    "language_info": {
      "name": "python",
      "version": "3.6.2",
      "mimetype": "text/x-python",
      "codemirror_mode": {
        "name": "ipython",
        "version": 3
      },
      "pygments_lexer": "ipython3",
      "nbconvert_exporter": "python",
      "file_extension": ".py"
    },
    "gist": {
      "id": "",
      "data": {
        "description": "nbs/keras_raw-vgg16.ipynb",
        "public": true
      }
    }
  },
  "nbformat": 4,
  "nbformat_minor": 2
}