Scene Classification(MobileNet).ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "unavailable-feeding",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-10-18 20:52:49.160471: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcudart.so.10.1\n"
     ]
    }
   ],
   "source": [
    "from tensorflow.keras.applications import MobileNet\n",
    "from tensorflow.keras.models import Sequential, Model\n",
    "from tensorflow.keras.layers import Dense, Dropout, Activation, Flatten, GlobalAveragePooling2D\n",
    "from tensorflow.keras.layers import Conv2D, MaxPooling2D, ZeroPadding2D\n",
    "from tensorflow.keras.layers import BatchNormalization\n",
    "from tensorflow.keras.preprocessing.image import ImageDataGenerator\n",
    "from tensorflow.keras.models import load_model\n",
    "\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "import wandb\n",
    "from wandb.keras import WandbCallback\n",
    "\n",
    "import os"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "furnished-cambodia",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-10-18 20:52:50.210196: I tensorflow/core/platform/cpu_feature_guard.cc:142] This TensorFlow binary is optimized with oneAPI Deep Neural Network Library (oneDNN) to use the following CPU instructions in performance-critical operations:  SSE4.1 SSE4.2 AVX AVX2 FMA\n",
      "To enable them in other operations, rebuild TensorFlow with the appropriate compiler flags.\n",
      "2021-10-18 20:52:50.211432: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcuda.so.1\n",
      "2021-10-18 20:52:50.248883: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:941] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero\n",
      "2021-10-18 20:52:50.249503: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1720] Found device 0 with properties: \n",
      "pciBusID: 0000:08:00.0 name: NVIDIA GeForce GTX 1660 SUPER computeCapability: 7.5\n",
      "coreClock: 1.86GHz coreCount: 22 deviceMemorySize: 5.80GiB deviceMemoryBandwidth: 312.97GiB/s\n",
      "2021-10-18 20:52:50.249518: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcudart.so.10.1\n",
      "2021-10-18 20:52:50.250492: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcublas.so.10\n",
      "2021-10-18 20:52:50.250514: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcublasLt.so.10\n",
      "2021-10-18 20:52:50.251478: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcufft.so.10\n",
      "2021-10-18 20:52:50.251619: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcurand.so.10\n",
      "2021-10-18 20:52:50.252499: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcusolver.so.10\n",
      "2021-10-18 20:52:50.253015: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcusparse.so.10\n",
      "2021-10-18 20:52:50.254930: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcudnn.so.7\n",
      "2021-10-18 20:52:50.255019: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:941] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero\n",
      "2021-10-18 20:52:50.255535: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:941] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero\n",
      "2021-10-18 20:52:50.255861: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1862] Adding visible gpu devices: 0\n",
      "2021-10-18 20:52:50.255883: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcudart.so.10.1\n",
      "2021-10-18 20:52:50.612949: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1261] Device interconnect StreamExecutor with strength 1 edge matrix:\n",
      "2021-10-18 20:52:50.612980: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1267]      0 \n",
      "2021-10-18 20:52:50.612985: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1280] 0:   N \n",
      "2021-10-18 20:52:50.613159: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:941] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero\n",
      "2021-10-18 20:52:50.613603: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:941] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero\n",
      "2021-10-18 20:52:50.613909: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:941] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero\n",
      "2021-10-18 20:52:50.614200: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1406] Created TensorFlow device (/job:localhost/replica:0/task:0/device:GPU:0 with 4753 MB memory) -> physical GPU (device: 0, name: NVIDIA GeForce GTX 1660 SUPER, pci bus id: 0000:08:00.0, compute capability: 7.5)\n",
      "2021-10-18 20:52:50.614350: I tensorflow/compiler/jit/xla_gpu_device.cc:99] Not creating XLA devices, tf_xla_enable_xla_devices not set\n"
     ]
    }
   ],
   "source": [
    "import tensorflow as tf\n",
    "\n",
    "os.environ['CUDA_VISIBLE_DEVICES'] = '0' # 초기화 할 GPU Number\n",
    "config = tf.compat.v1.ConfigProto()\n",
    "config.gpu_options.per_process_gpu_memory_fraction = 0.8\n",
    "sess = tf.compat.v1.Session(config=config)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "handled-demographic",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-10-18 20:52:50.631342: I tensorflow/compiler/jit/xla_cpu_device.cc:41] Not creating XLA devices, tf_xla_enable_xla_devices not set\n",
      "2021-10-18 20:52:50.631511: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:941] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero\n",
      "2021-10-18 20:52:50.632343: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1720] Found device 0 with properties: \n",
      "pciBusID: 0000:08:00.0 name: NVIDIA GeForce GTX 1660 SUPER computeCapability: 7.5\n",
      "coreClock: 1.86GHz coreCount: 22 deviceMemorySize: 5.80GiB deviceMemoryBandwidth: 312.97GiB/s\n",
      "2021-10-18 20:52:50.632380: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcudart.so.10.1\n",
      "2021-10-18 20:52:50.632404: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcublas.so.10\n",
      "2021-10-18 20:52:50.632415: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcublasLt.so.10\n",
      "2021-10-18 20:52:50.632425: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcufft.so.10\n",
      "2021-10-18 20:52:50.632433: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcurand.so.10\n",
      "2021-10-18 20:52:50.632440: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcusolver.so.10\n",
      "2021-10-18 20:52:50.632446: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcusparse.so.10\n",
      "2021-10-18 20:52:50.632453: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcudnn.so.7\n",
      "2021-10-18 20:52:50.632491: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:941] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero\n",
      "2021-10-18 20:52:50.632803: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:941] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero\n",
      "2021-10-18 20:52:50.633078: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1862] Adding visible gpu devices: 0\n",
      "2021-10-18 20:52:50.633227: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:941] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero\n",
      "2021-10-18 20:52:50.633579: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1720] Found device 0 with properties: \n",
      "pciBusID: 0000:08:00.0 name: NVIDIA GeForce GTX 1660 SUPER computeCapability: 7.5\n",
      "coreClock: 1.86GHz coreCount: 22 deviceMemorySize: 5.80GiB deviceMemoryBandwidth: 312.97GiB/s\n",
      "2021-10-18 20:52:50.633588: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcudart.so.10.1\n",
      "2021-10-18 20:52:50.633596: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcublas.so.10\n",
      "2021-10-18 20:52:50.633602: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcublasLt.so.10\n",
      "2021-10-18 20:52:50.633608: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcufft.so.10\n",
      "2021-10-18 20:52:50.633613: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcurand.so.10\n",
      "2021-10-18 20:52:50.633619: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcusolver.so.10\n",
      "2021-10-18 20:52:50.633625: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcusparse.so.10\n",
      "2021-10-18 20:52:50.633631: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcudnn.so.7\n",
      "2021-10-18 20:52:50.633661: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:941] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero\n",
      "2021-10-18 20:52:50.633987: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:941] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero\n",
      "2021-10-18 20:52:50.634322: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1862] Adding visible gpu devices: 0\n",
      "2021-10-18 20:52:50.634343: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1261] Device interconnect StreamExecutor with strength 1 edge matrix:\n",
      "2021-10-18 20:52:50.634347: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1267]      0 \n",
      "2021-10-18 20:52:50.634351: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1280] 0:   N \n",
      "2021-10-18 20:52:50.634405: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:941] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero\n",
      "2021-10-18 20:52:50.634890: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:941] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero\n",
      "2021-10-18 20:52:50.636087: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1406] Created TensorFlow device (/job:localhost/replica:0/task:0/device:GPU:0 with 4753 MB memory) -> physical GPU (device: 0, name: NVIDIA GeForce GTX 1660 SUPER, pci bus id: 0000:08:00.0, compute capability: 7.5)\n",
      "2021-10-18 20:52:50.636097: I tensorflow/compiler/jit/xla_gpu_device.cc:99] Not creating XLA devices, tf_xla_enable_xla_devices not set\n"
     ]
    }
   ],
   "source": [
    "img_rows , img_cols = 224, 224 # MobileNet input size\n",
    "\n",
    "# imagenet pretrained된 가중치 가져옴\n",
    "MobileNet = MobileNet(weights='imagenet', include_top=False, input_shape=(img_rows, img_cols, 3))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "charged-textbook",
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0 InputLayer True\n",
      "1 Conv2D True\n",
      "2 BatchNormalization True\n",
      "3 ReLU True\n",
      "4 DepthwiseConv2D True\n",
      "5 BatchNormalization True\n",
      "6 ReLU True\n",
      "7 Conv2D True\n",
      "8 BatchNormalization True\n",
      "9 ReLU True\n",
      "10 ZeroPadding2D True\n",
      "11 DepthwiseConv2D True\n",
      "12 BatchNormalization True\n",
      "13 ReLU True\n",
      "14 Conv2D True\n",
      "15 BatchNormalization True\n",
      "16 ReLU True\n",
      "17 DepthwiseConv2D True\n",
      "18 BatchNormalization True\n",
      "19 ReLU True\n",
      "20 Conv2D True\n",
      "21 BatchNormalization True\n",
      "22 ReLU True\n",
      "23 ZeroPadding2D True\n",
      "24 DepthwiseConv2D True\n",
      "25 BatchNormalization True\n",
      "26 ReLU True\n",
      "27 Conv2D True\n",
      "28 BatchNormalization True\n",
      "29 ReLU True\n",
      "30 DepthwiseConv2D True\n",
      "31 BatchNormalization True\n",
      "32 ReLU True\n",
      "33 Conv2D True\n",
      "34 BatchNormalization True\n",
      "35 ReLU True\n",
      "36 ZeroPadding2D True\n",
      "37 DepthwiseConv2D True\n",
      "38 BatchNormalization True\n",
      "39 ReLU True\n",
      "40 Conv2D True\n",
      "41 BatchNormalization True\n",
      "42 ReLU True\n",
      "43 DepthwiseConv2D True\n",
      "44 BatchNormalization True\n",
      "45 ReLU True\n",
      "46 Conv2D True\n",
      "47 BatchNormalization True\n",
      "48 ReLU True\n",
      "49 DepthwiseConv2D True\n",
      "50 BatchNormalization True\n",
      "51 ReLU True\n",
      "52 Conv2D True\n",
      "53 BatchNormalization True\n",
      "54 ReLU True\n",
      "55 DepthwiseConv2D True\n",
      "56 BatchNormalization True\n",
      "57 ReLU True\n",
      "58 Conv2D True\n",
      "59 BatchNormalization True\n",
      "60 ReLU True\n",
      "61 DepthwiseConv2D True\n",
      "62 BatchNormalization True\n",
      "63 ReLU True\n",
      "64 Conv2D True\n",
      "65 BatchNormalization True\n",
      "66 ReLU True\n",
      "67 DepthwiseConv2D True\n",
      "68 BatchNormalization True\n",
      "69 ReLU True\n",
      "70 Conv2D True\n",
      "71 BatchNormalization True\n",
      "72 ReLU True\n",
      "73 ZeroPadding2D True\n",
      "74 DepthwiseConv2D True\n",
      "75 BatchNormalization True\n",
      "76 ReLU True\n",
      "77 Conv2D True\n",
      "78 BatchNormalization True\n",
      "79 ReLU True\n",
      "80 DepthwiseConv2D True\n",
      "81 BatchNormalization True\n",
      "82 ReLU True\n",
      "83 Conv2D True\n",
      "84 BatchNormalization True\n",
      "85 ReLU True\n"
     ]
    }
   ],
   "source": [
    "for layer in MobileNet.layers:\n",
    "    layer.trainable = True\n",
    "    \n",
    "for (i, layer) in enumerate(MobileNet.layers):\n",
    "    print(str(i), layer.__class__.__name__, layer.trainable)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "alert-citizenship",
   "metadata": {},
   "outputs": [],
   "source": [
    "def addTopModelMobileNet(bottom_model, num_classes):\n",
    "    top_model = bottom_model.output\n",
    "    top_model = GlobalAveragePooling2D()(top_model)\n",
    "    top_model = Dense(1024, activation='relu')(top_model)\n",
    "    top_model = Dense(1024, activation='relu')(top_model)\n",
    "    top_model = Dense(512, activation='relu')(top_model)\n",
    "    top_model = Dense(num_classes, activation='softmax')(top_model)\n",
    "    \n",
    "    return top_model\n",
    "    "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "imported-outdoors",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Found 2446 images belonging to 2 classes.\n",
      "Found 613 images belonging to 2 classes.\n",
      "Found 833 images belonging to 2 classes.\n"
     ]
    }
   ],
   "source": [
    "train_data_dir = 'paper_dataset/train'\n",
    "validation_data_dir = 'paper_dataset/validation'\n",
    "test_data_dir = 'paper_dataset/test'\n",
    "\n",
    "train_datagen = ImageDataGenerator(\n",
    "                    rescale=1./255,\n",
    "#                     rotation_range=30,\n",
    "#                     width_shift_range=0.3,\n",
    "#                     height_shift_range=0.3,\n",
    "#                     fill_mode='nearest'\n",
    "                    )\n",
    "\n",
    "validation_datagen = ImageDataGenerator(rescale=1./255)\n",
    "\n",
    "test_datagen = ImageDataGenerator(rescale=1./255)\n",
    "\n",
    "batch_size = 32\n",
    "\n",
    "train_generator = train_datagen.flow_from_directory(\n",
    "train_data_dir,\n",
    "target_size=(img_rows, img_cols),\n",
    "batch_size=batch_size,\n",
    "class_mode='categorical',\n",
    ")\n",
    "\n",
    "validation_generator = validation_datagen.flow_from_directory(\n",
    "validation_data_dir,\n",
    "target_size=(img_rows, img_cols),\n",
    "batch_size=batch_size,\n",
    "class_mode='categorical'\n",
    ")\n",
    "\n",
    "test_generator = test_datagen.flow_from_directory(\n",
    "    test_data_dir,\n",
    "    target_size=(img_rows, img_cols),\n",
    "    batch_size=batch_size,\n",
    "    class_mode='categorical'\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "ruled-locator",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"model\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "input_1 (InputLayer)         [(None, 224, 224, 3)]     0         \n",
      "_________________________________________________________________\n",
      "conv1 (Conv2D)               (None, 112, 112, 32)      864       \n",
      "_________________________________________________________________\n",
      "conv1_bn (BatchNormalization (None, 112, 112, 32)      128       \n",
      "_________________________________________________________________\n",
      "conv1_relu (ReLU)            (None, 112, 112, 32)      0         \n",
      "_________________________________________________________________\n",
      "conv_dw_1 (DepthwiseConv2D)  (None, 112, 112, 32)      288       \n",
      "_________________________________________________________________\n",
      "conv_dw_1_bn (BatchNormaliza (None, 112, 112, 32)      128       \n",
      "_________________________________________________________________\n",
      "conv_dw_1_relu (ReLU)        (None, 112, 112, 32)      0         \n",
      "_________________________________________________________________\n",
      "conv_pw_1 (Conv2D)           (None, 112, 112, 64)      2048      \n",
      "_________________________________________________________________\n",
      "conv_pw_1_bn (BatchNormaliza (None, 112, 112, 64)      256       \n",
      "_________________________________________________________________\n",
      "conv_pw_1_relu (ReLU)        (None, 112, 112, 64)      0         \n",
      "_________________________________________________________________\n",
      "conv_pad_2 (ZeroPadding2D)   (None, 113, 113, 64)      0         \n",
      "_________________________________________________________________\n",
      "conv_dw_2 (DepthwiseConv2D)  (None, 56, 56, 64)        576       \n",
      "_________________________________________________________________\n",
      "conv_dw_2_bn (BatchNormaliza (None, 56, 56, 64)        256       \n",
      "_________________________________________________________________\n",
      "conv_dw_2_relu (ReLU)        (None, 56, 56, 64)        0         \n",
      "_________________________________________________________________\n",
      "conv_pw_2 (Conv2D)           (None, 56, 56, 128)       8192      \n",
      "_________________________________________________________________\n",
      "conv_pw_2_bn (BatchNormaliza (None, 56, 56, 128)       512       \n",
      "_________________________________________________________________\n",
      "conv_pw_2_relu (ReLU)        (None, 56, 56, 128)       0         \n",
      "_________________________________________________________________\n",
      "conv_dw_3 (DepthwiseConv2D)  (None, 56, 56, 128)       1152      \n",
      "_________________________________________________________________\n",
      "conv_dw_3_bn (BatchNormaliza (None, 56, 56, 128)       512       \n",
      "_________________________________________________________________\n",
      "conv_dw_3_relu (ReLU)        (None, 56, 56, 128)       0         \n",
      "_________________________________________________________________\n",
      "conv_pw_3 (Conv2D)           (None, 56, 56, 128)       16384     \n",
      "_________________________________________________________________\n",
      "conv_pw_3_bn (BatchNormaliza (None, 56, 56, 128)       512       \n",
      "_________________________________________________________________\n",
      "conv_pw_3_relu (ReLU)        (None, 56, 56, 128)       0         \n",
      "_________________________________________________________________\n",
      "conv_pad_4 (ZeroPadding2D)   (None, 57, 57, 128)       0         \n",
      "_________________________________________________________________\n",
      "conv_dw_4 (DepthwiseConv2D)  (None, 28, 28, 128)       1152      \n",
      "_________________________________________________________________\n",
      "conv_dw_4_bn (BatchNormaliza (None, 28, 28, 128)       512       \n",
      "_________________________________________________________________\n",
      "conv_dw_4_relu (ReLU)        (None, 28, 28, 128)       0         \n",
      "_________________________________________________________________\n",
      "conv_pw_4 (Conv2D)           (None, 28, 28, 256)       32768     \n",
      "_________________________________________________________________\n",
      "conv_pw_4_bn (BatchNormaliza (None, 28, 28, 256)       1024      \n",
      "_________________________________________________________________\n",
      "conv_pw_4_relu (ReLU)        (None, 28, 28, 256)       0         \n",
      "_________________________________________________________________\n",
      "conv_dw_5 (DepthwiseConv2D)  (None, 28, 28, 256)       2304      \n",
      "_________________________________________________________________\n",
      "conv_dw_5_bn (BatchNormaliza (None, 28, 28, 256)       1024      \n",
      "_________________________________________________________________\n",
      "conv_dw_5_relu (ReLU)        (None, 28, 28, 256)       0         \n",
      "_________________________________________________________________\n",
      "conv_pw_5 (Conv2D)           (None, 28, 28, 256)       65536     \n",
      "_________________________________________________________________\n",
      "conv_pw_5_bn (BatchNormaliza (None, 28, 28, 256)       1024      \n",
      "_________________________________________________________________\n",
      "conv_pw_5_relu (ReLU)        (None, 28, 28, 256)       0         \n",
      "_________________________________________________________________\n",
      "conv_pad_6 (ZeroPadding2D)   (None, 29, 29, 256)       0         \n",
      "_________________________________________________________________\n",
      "conv_dw_6 (DepthwiseConv2D)  (None, 14, 14, 256)       2304      \n",
      "_________________________________________________________________\n",
      "conv_dw_6_bn (BatchNormaliza (None, 14, 14, 256)       1024      \n",
      "_________________________________________________________________\n",
      "conv_dw_6_relu (ReLU)        (None, 14, 14, 256)       0         \n",
      "_________________________________________________________________\n",
      "conv_pw_6 (Conv2D)           (None, 14, 14, 512)       131072    \n",
      "_________________________________________________________________\n",
      "conv_pw_6_bn (BatchNormaliza (None, 14, 14, 512)       2048      \n",
      "_________________________________________________________________\n",
      "conv_pw_6_relu (ReLU)        (None, 14, 14, 512)       0         \n",
      "_________________________________________________________________\n",
      "conv_dw_7 (DepthwiseConv2D)  (None, 14, 14, 512)       4608      \n",
      "_________________________________________________________________\n",
      "conv_dw_7_bn (BatchNormaliza (None, 14, 14, 512)       2048      \n",
      "_________________________________________________________________\n",
      "conv_dw_7_relu (ReLU)        (None, 14, 14, 512)       0         \n",
      "_________________________________________________________________\n",
      "conv_pw_7 (Conv2D)           (None, 14, 14, 512)       262144    \n",
      "_________________________________________________________________\n",
      "conv_pw_7_bn (BatchNormaliza (None, 14, 14, 512)       2048      \n",
      "_________________________________________________________________\n",
      "conv_pw_7_relu (ReLU)        (None, 14, 14, 512)       0         \n",
      "_________________________________________________________________\n",
      "conv_dw_8 (DepthwiseConv2D)  (None, 14, 14, 512)       4608      \n",
      "_________________________________________________________________\n",
      "conv_dw_8_bn (BatchNormaliza (None, 14, 14, 512)       2048      \n",
      "_________________________________________________________________\n",
      "conv_dw_8_relu (ReLU)        (None, 14, 14, 512)       0         \n",
      "_________________________________________________________________\n",
      "conv_pw_8 (Conv2D)           (None, 14, 14, 512)       262144    \n",
      "_________________________________________________________________\n",
      "conv_pw_8_bn (BatchNormaliza (None, 14, 14, 512)       2048      \n",
      "_________________________________________________________________\n",
      "conv_pw_8_relu (ReLU)        (None, 14, 14, 512)       0         \n",
      "_________________________________________________________________\n",
      "conv_dw_9 (DepthwiseConv2D)  (None, 14, 14, 512)       4608      \n",
      "_________________________________________________________________\n",
      "conv_dw_9_bn (BatchNormaliza (None, 14, 14, 512)       2048      \n",
      "_________________________________________________________________\n",
      "conv_dw_9_relu (ReLU)        (None, 14, 14, 512)       0         \n",
      "_________________________________________________________________\n",
      "conv_pw_9 (Conv2D)           (None, 14, 14, 512)       262144    \n",
      "_________________________________________________________________\n",
      "conv_pw_9_bn (BatchNormaliza (None, 14, 14, 512)       2048      \n",
      "_________________________________________________________________\n",
      "conv_pw_9_relu (ReLU)        (None, 14, 14, 512)       0         \n",
      "_________________________________________________________________\n",
      "conv_dw_10 (DepthwiseConv2D) (None, 14, 14, 512)       4608      \n",
      "_________________________________________________________________\n",
      "conv_dw_10_bn (BatchNormaliz (None, 14, 14, 512)       2048      \n",
      "_________________________________________________________________\n",
      "conv_dw_10_relu (ReLU)       (None, 14, 14, 512)       0         \n",
      "_________________________________________________________________\n",
      "conv_pw_10 (Conv2D)          (None, 14, 14, 512)       262144    \n",
      "_________________________________________________________________\n",
      "conv_pw_10_bn (BatchNormaliz (None, 14, 14, 512)       2048      \n",
      "_________________________________________________________________\n",
      "conv_pw_10_relu (ReLU)       (None, 14, 14, 512)       0         \n",
      "_________________________________________________________________\n",
      "conv_dw_11 (DepthwiseConv2D) (None, 14, 14, 512)       4608      \n",
      "_________________________________________________________________\n",
      "conv_dw_11_bn (BatchNormaliz (None, 14, 14, 512)       2048      \n",
      "_________________________________________________________________\n",
      "conv_dw_11_relu (ReLU)       (None, 14, 14, 512)       0         \n",
      "_________________________________________________________________\n",
      "conv_pw_11 (Conv2D)          (None, 14, 14, 512)       262144    \n",
      "_________________________________________________________________\n",
      "conv_pw_11_bn (BatchNormaliz (None, 14, 14, 512)       2048      \n",
      "_________________________________________________________________\n",
      "conv_pw_11_relu (ReLU)       (None, 14, 14, 512)       0         \n",
      "_________________________________________________________________\n",
      "conv_pad_12 (ZeroPadding2D)  (None, 15, 15, 512)       0         \n",
      "_________________________________________________________________\n",
      "conv_dw_12 (DepthwiseConv2D) (None, 7, 7, 512)         4608      \n",
      "_________________________________________________________________\n",
      "conv_dw_12_bn (BatchNormaliz (None, 7, 7, 512)         2048      \n",
      "_________________________________________________________________\n",
      "conv_dw_12_relu (ReLU)       (None, 7, 7, 512)         0         \n",
      "_________________________________________________________________\n",
      "conv_pw_12 (Conv2D)          (None, 7, 7, 1024)        524288    \n",
      "_________________________________________________________________\n",
      "conv_pw_12_bn (BatchNormaliz (None, 7, 7, 1024)        4096      \n",
      "_________________________________________________________________\n",
      "conv_pw_12_relu (ReLU)       (None, 7, 7, 1024)        0         \n",
      "_________________________________________________________________\n",
      "conv_dw_13 (DepthwiseConv2D) (None, 7, 7, 1024)        9216      \n",
      "_________________________________________________________________\n",
      "conv_dw_13_bn (BatchNormaliz (None, 7, 7, 1024)        4096      \n",
      "_________________________________________________________________\n",
      "conv_dw_13_relu (ReLU)       (None, 7, 7, 1024)        0         \n",
      "_________________________________________________________________\n",
      "conv_pw_13 (Conv2D)          (None, 7, 7, 1024)        1048576   \n",
      "_________________________________________________________________\n",
      "conv_pw_13_bn (BatchNormaliz (None, 7, 7, 1024)        4096      \n",
      "_________________________________________________________________\n",
      "conv_pw_13_relu (ReLU)       (None, 7, 7, 1024)        0         \n",
      "_________________________________________________________________\n",
      "global_average_pooling2d (Gl (None, 1024)              0         \n",
      "_________________________________________________________________\n",
      "dense (Dense)                (None, 1024)              1049600   \n",
      "_________________________________________________________________\n",
      "dense_1 (Dense)              (None, 1024)              1049600   \n",
      "_________________________________________________________________\n",
      "dense_2 (Dense)              (None, 512)               524800    \n",
      "_________________________________________________________________\n",
      "dense_3 (Dense)              (None, 2)                 1026      \n",
      "=================================================================\n",
      "Total params: 5,853,890\n",
      "Trainable params: 5,832,002\n",
      "Non-trainable params: 21,888\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "num_classes = train_generator.num_classes\n",
    "FC_Head = addTopModelMobileNet(MobileNet, num_classes)\n",
    "\n",
    "model = Model(inputs=MobileNet.input, outputs=FC_Head)\n",
    "\n",
    "model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "exposed-language",
   "metadata": {},
   "outputs": [],
   "source": [
    "from tensorflow.keras.optimizers import RMSprop, Adam, SGD\n",
    "from tensorflow.keras.callbacks import ModelCheckpoint, EarlyStopping, ReduceLROnPlateau, TensorBoard\n",
    "from datetime import datetime"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "id": "sensitive-sarah",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Failed to detect the name of this notebook, you can set it manually with the WANDB_NOTEBOOK_NAME environment variable to enable code saving.\n",
      "\u001b[34m\u001b[1mwandb\u001b[0m: Currently logged in as: \u001b[33myhkim\u001b[0m (use `wandb login --relogin` to force relogin)\n",
      "/home/yongho/anaconda3/envs/tensor/lib/python3.7/site-packages/IPython/html.py:14: ShimWarning: The `IPython.html` package has been deprecated since IPython 4.0. You should import from `notebook` instead. `IPython.html.widgets` has moved to `ipywidgets`.\n",
      "  \"`IPython.html.widgets` has moved to `ipywidgets`.\", ShimWarning)\n",
      "2021-10-18 20:52:52.599112: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcudart.so.10.1\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "\n",
       "                    Syncing run <strong><a href=\"https://wandb.ai/yhkim/KICS_FALL/runs/1judgr9x\" target=\"_blank\">genial-sponge-3</a></strong> to <a href=\"https://wandb.ai/yhkim/KICS_FALL\" target=\"_blank\">Weights & Biases</a> (<a href=\"https://docs.wandb.com/integrations/jupyter.html\" target=\"_blank\">docs</a>).<br/>\n",
       "\n",
       "                "
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "wandb.init(project='KICS_FALL', entity='yhkim')\n",
    "config = wandb.config\n",
    "\n",
    "config.learning_rate = 0.01\n",
    "config.momentum = 0.937\n",
    "config.nesterov = True\n",
    "config.epochs =100"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "id": "economic-thanksgiving",
   "metadata": {},
   "outputs": [],
   "source": [
    "checkpoint = ModelCheckpoint(\n",
    "    'scene_clasification_MobileNet.h5',\n",
    "    monitor='val_loss',\n",
    "    mode='min',\n",
    "    save_best_only=True,\n",
    "    verbose=1\n",
    ")\n",
    "\n",
    "earlystop = EarlyStopping(\n",
    "    monitor='val_loss',\n",
    "    min_delta=0,\n",
    "    patience=10,\n",
    "    verbose=1,\n",
    "    restore_best_weights=True\n",
    ")\n",
    "\n",
    "learning_rate_reduction = ReduceLROnPlateau(monitor='val_accuracy',\n",
    "                                           patience=5,\n",
    "                                           verbose=1,\n",
    "                                           factor=0.2,\n",
    "                                           min_lr=0.0001)\n",
    "\n",
    "# tensorboard = TensorBoard(log_dir=f'./logs/{datetime.now().strftime(\"%Y%m%d-%H%M%S\")}')\n",
    "\n",
    "callbacks = [checkpoint, earlystop, WandbCallback(data_type='image', labels=['not_pitching', 'pitching'])]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "id": "latin-vegetation",
   "metadata": {},
   "outputs": [],
   "source": [
    "model.compile(loss='categorical_crossentropy',\n",
    "             optimizer=SGD(lr=config.learning_rate, momentum=config.momentum,\n",
    "                           nesterov=config.nesterov),\n",
    "             metrics=['accuracy']\n",
    "             )\n",
    "\n",
    "nb_train_samples = len(train_generator.classes)\n",
    "nb_validation_samples = len(validation_generator.classes)\n",
    "\n",
    "epochs = config.epochs"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "id": "constant-iraqi",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/yongho/anaconda3/envs/tensor/lib/python3.7/site-packages/tensorflow/python/keras/engine/training.py:1844: UserWarning: `Model.fit_generator` is deprecated and will be removed in a future version. Please use `Model.fit`, which supports generators.\n",
      "  warnings.warn('`Model.fit_generator` is deprecated and '\n",
      "2021-10-18 20:52:56.391951: I tensorflow/compiler/mlir/mlir_graph_optimization_pass.cc:116] None of the MLIR optimization passes are enabled (registered 2)\n",
      "2021-10-18 20:52:56.408394: I tensorflow/core/platform/profile_utils/cpu_utils.cc:112] CPU Frequency: 3593330000 Hz\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/100\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-10-18 20:52:57.487140: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcublas.so.10\n",
      "2021-10-18 20:52:57.774779: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcudnn.so.7\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "76/76 [==============================] - 29s 351ms/step - loss: 0.3034 - accuracy: 0.8565 - val_loss: 0.1058 - val_accuracy: 0.9720\n",
      "\n",
      "Epoch 00001: val_loss improved from inf to 0.10576, saving model to scene_clasification_MobileNet.h5\n",
      "Epoch 2/100\n",
      "76/76 [==============================] - 26s 340ms/step - loss: 0.0535 - accuracy: 0.9832 - val_loss: 0.0494 - val_accuracy: 0.9852\n",
      "\n",
      "Epoch 00002: val_loss improved from 0.10576 to 0.04944, saving model to scene_clasification_MobileNet.h5\n",
      "Epoch 3/100\n",
      "76/76 [==============================] - 26s 342ms/step - loss: 0.0283 - accuracy: 0.9909 - val_loss: 0.0577 - val_accuracy: 0.9819\n",
      "\n",
      "Epoch 00003: val_loss did not improve from 0.04944\n",
      "Epoch 4/100\n",
      "76/76 [==============================] - 26s 343ms/step - loss: 0.0130 - accuracy: 0.9966 - val_loss: 0.1033 - val_accuracy: 0.9786\n",
      "\n",
      "Epoch 00004: val_loss did not improve from 0.04944\n",
      "Epoch 5/100\n",
      "76/76 [==============================] - 26s 345ms/step - loss: 0.0111 - accuracy: 0.9972 - val_loss: 0.0793 - val_accuracy: 0.9704\n",
      "\n",
      "Epoch 00005: val_loss did not improve from 0.04944\n",
      "Epoch 6/100\n",
      "76/76 [==============================] - 27s 348ms/step - loss: 0.0233 - accuracy: 0.9917 - val_loss: 0.2922 - val_accuracy: 0.9523\n",
      "\n",
      "Epoch 00006: val_loss did not improve from 0.04944\n",
      "Epoch 7/100\n",
      "76/76 [==============================] - 27s 348ms/step - loss: 0.0050 - accuracy: 0.9978 - val_loss: 0.0645 - val_accuracy: 0.9803\n",
      "\n",
      "Epoch 00007: val_loss did not improve from 0.04944\n",
      "Epoch 8/100\n",
      "76/76 [==============================] - 26s 345ms/step - loss: 0.0174 - accuracy: 0.9934 - val_loss: 0.1061 - val_accuracy: 0.9803\n",
      "\n",
      "Epoch 00008: val_loss did not improve from 0.04944\n",
      "Epoch 9/100\n",
      "76/76 [==============================] - 26s 345ms/step - loss: 0.0170 - accuracy: 0.9956 - val_loss: 0.1031 - val_accuracy: 0.9720\n",
      "\n",
      "Epoch 00009: val_loss did not improve from 0.04944\n",
      "Epoch 10/100\n",
      "76/76 [==============================] - 26s 344ms/step - loss: 0.0154 - accuracy: 0.9949 - val_loss: 0.0850 - val_accuracy: 0.9770\n",
      "\n",
      "Epoch 00010: val_loss did not improve from 0.04944\n",
      "Epoch 11/100\n",
      "76/76 [==============================] - 26s 345ms/step - loss: 0.0148 - accuracy: 0.9953 - val_loss: 0.0946 - val_accuracy: 0.9786\n",
      "\n",
      "Epoch 00011: val_loss did not improve from 0.04944\n",
      "Epoch 12/100\n",
      "76/76 [==============================] - 26s 345ms/step - loss: 0.0028 - accuracy: 0.9984 - val_loss: 0.1165 - val_accuracy: 0.9786\n",
      "\n",
      "Epoch 00012: val_loss did not improve from 0.04944\n",
      "Restoring model weights from the end of the best epoch.\n",
      "Epoch 00012: early stopping\n"
     ]
    }
   ],
   "source": [
    "history = model.fit_generator(train_generator,\n",
    "                              steps_per_epoch=nb_train_samples//batch_size,\n",
    "                              validation_data=validation_generator,\n",
    "                              validation_steps=nb_validation_samples//batch_size,\n",
    "                              epochs=config.epochs,\n",
    "                              callbacks=callbacks,\n",
    "                              )"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "id": "announced-solomon",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "history = model.history\n",
    "\n",
    "plt.plot(history.history['accuracy'])\n",
    "plt.plot(history.history['val_accuracy'])\n",
    "\n",
    "plt.title('Model accuracy')\n",
    "plt.xlabel('Epoch')\n",
    "plt.ylabel('Accuracy')\n",
    "plt.legend(['Train', 'Validation'], loc='upper left')\n",
    "plt.show()\n",
    "\n",
    "plt.plot(history.history['loss'])\n",
    "plt.plot(history.history['val_loss'])\n",
    "plt.title('Model loss')\n",
    "plt.xlabel('Epoch')\n",
    "plt.legend(['Train', 'Validation'], loc='upper left')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "id": "radical-shooting",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-10-18 20:58:30.431489: W tensorflow/python/util/util.cc:348] Sets are not currently considered sequences, but this may change in the future, so consider avoiding using them.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Assets written to: paper_data/assets\n"
     ]
    }
   ],
   "source": [
    "model.save('paper_data')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "id": "enhanced-isaac",
   "metadata": {},
   "outputs": [],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "import cv2\n",
    "import numpy as np\n",
    "import random\n",
    "from sklearn.metrics import confusion_matrix, classification_report, plot_confusion_matrix, accuracy_score\n",
    "import sklearn"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "id": "trying-nancy",
   "metadata": {},
   "outputs": [],
   "source": [
    "model = load_model('./paper_data')\n",
    "classes = os.listdir('./paper_dataset/train')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "id": "discrete-detector",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Found 833 images belonging to 2 classes.\n"
     ]
    }
   ],
   "source": [
    "img_rows, img_cols = 224, 224\n",
    "batch_size = 64\n",
    "\n",
    "test_data_dir = 'paper_dataset/test'\n",
    "test_datagen = ImageDataGenerator(rescale=1./255)\n",
    "test_generator = test_datagen.flow_from_directory(\n",
    "    test_data_dir,\n",
    "    target_size=(img_rows, img_cols),\n",
    "    batch_size=batch_size,\n",
    "    class_mode='categorical',\n",
    "    shuffle=False\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "id": "threatened-surgeon",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/yongho/anaconda3/envs/tensor/lib/python3.7/site-packages/tensorflow/python/keras/engine/training.py:1905: UserWarning: `Model.predict_generator` is deprecated and will be removed in a future version. Please use `Model.predict`, which supports generators.\n",
      "  warnings.warn('`Model.predict_generator` is deprecated and '\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "14/14 [==============================] - 4s 238ms/step\n"
     ]
    }
   ],
   "source": [
    "test_generator.reset()\n",
    "Y_pred = model.predict_generator(test_generator, verbose=1)\n",
    "classes = test_generator.classes[test_generator.index_array]\n",
    "y_pred = np.argmax(Y_pred, axis=-1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "id": "liberal-belle",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.9015606242496998\n"
     ]
    }
   ],
   "source": [
    "print(sum(y_pred==classes)/833)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "auburn-toner",
   "metadata": {},
   "source": [
    "## Scene Classification"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "id": "proprietary-conversation",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "-- Confusion Matrix --\n",
      "[[656   0]\n",
      " [ 82  95]]\n"
     ]
    }
   ],
   "source": [
    "print('-- Confusion Matrix --')\n",
    "print(confusion_matrix(test_generator.classes[test_generator.index_array], y_pred))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "id": "fatal-manitoba",
   "metadata": {},
   "outputs": [],
   "source": [
    "class_names = ['not_pitching', 'pitching']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "id": "suburban-scale",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "              precision    recall  f1-score   support\n",
      "\n",
      "not_pitching       0.89      1.00      0.94       656\n",
      "    pitching       1.00      0.54      0.70       177\n",
      "\n",
      "    accuracy                           0.90       833\n",
      "   macro avg       0.94      0.77      0.82       833\n",
      "weighted avg       0.91      0.90      0.89       833\n",
      "\n"
     ]
    }
   ],
   "source": [
    "print(classification_report(test_generator.classes[test_generator.index_array], y_pred, target_names=class_names))"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "local-seller",
   "metadata": {},
   "source": [
    "## Object Detection"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "id": "aware-punch",
   "metadata": {},
   "outputs": [
    {
     "ename": "FileNotFoundError",
     "evalue": "[Errno 2] No such file or directory: 'od_result'",
     "output_type": "error",
     "traceback": [
      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[0;31mFileNotFoundError\u001b[0m                         Traceback (most recent call last)",
      "\u001b[0;32m/tmp/ipykernel_4163/1050623370.py\u001b[0m in \u001b[0;36m<module>\u001b[0;34m\u001b[0m\n\u001b[1;32m      5\u001b[0m     \u001b[0mod_true\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0mint\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mx\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mreplace\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m'.'\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m'_'\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0msplit\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m'_'\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;36m1\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m]\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0;36m1\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      6\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m----> 7\u001b[0;31m \u001b[0;32mfor\u001b[0m \u001b[0mx\u001b[0m \u001b[0;32min\u001b[0m \u001b[0mos\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mlistdir\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m'od_result'\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m      8\u001b[0m     \u001b[0mod_predict\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0mint\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mx\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mreplace\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m'.'\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m'_'\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0msplit\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m'_'\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;36m1\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m]\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0;36m1\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;31mFileNotFoundError\u001b[0m: [Errno 2] No such file or directory: 'od_result'"
     ]
    }
   ],
   "source": [
    "od_true = np.zeros((833,), dtype=int)\n",
    "od_predict = np.zeros((833,), dtype=int)\n",
    "\n",
    "for x in os.listdir('paper_dataset/test/pitching'):\n",
    "    od_true[int(x.replace('.','_').split('_')[1])] = 1\n",
    "\n",
    "for x in os.listdir('od_result'):\n",
    "    od_predict[int(x.replace('.','_').split('_')[1])] = 1"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "convenient-passage",
   "metadata": {},
   "outputs": [],
   "source": [
    "print('-- Confusion Matrix --')\n",
    "print(confusion_matrix(od_true, od_predict))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "revised-agenda",
   "metadata": {},
   "outputs": [],
   "source": [
    "print(classification_report(od_true, od_predict, target_names=class_names))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "polished-invalid",
   "metadata": {},
   "outputs": [],
   "source": [
    "accuracy_score(od_true, od_predict)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "ecological-jacket",
   "metadata": {},
   "outputs": [],
   "source": [
    "scores = model.evaluate_generator(test_generator)\n",
    "print(\"%s: %.2f%%\" %(model.metrics_names[1], scores[1]*100))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "arabic-punishment",
   "metadata": {},
   "outputs": [],
   "source": [
    "model.metrics_names"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "dental-modem",
   "metadata": {},
   "outputs": [],
   "source": [
    "# # random_img = random.sample(os.listdir('./paper_dataset/test/not_pitching/'), 1)[0]\n",
    "# img = cv2.imread(f\"./paper_dataset/test/pitching/frame_515.jpg\")\n",
    "# resize_img = cv2.resize(img.astype('float32') / 255, (224, 224))\n",
    "# # print(random_img)\n",
    "\n",
    "# output = model.predict(resize_img[np.newaxis])\n",
    "\n",
    "# plt.title(f\"True : not_pitching       Predict : {'not pitching' if np.argmax(output) == 0 else 'pitching'}\")\n",
    "# plt.imshow(cv2.cvtColor(img, cv2.COLOR_BGR2RGB))\n",
    "# plt.savefig('Predict not_pitching scene', format='jpg')\n",
    "# # cv2.imwrite('Predict pitching scene.jpg', img)\n",
    "# # plt.imsave('Predict pitching scene.jpg', cv2.cvtColor(img, cv2.COLOR_BGR2RGB))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "unauthorized-perth",
   "metadata": {},
   "outputs": [],
   "source": [
    "# outputs = model.predict_generator(test_generator, verbose=1)\n",
    "# np.set_printoptions(formatter={'float': lambda x: \"{0:0.3f}\".format(x)})\n",
    "# print(test_generator.class_indices)\n",
    "# print(outputs)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "vital-brush",
   "metadata": {},
   "outputs": [],
   "source": [
    "# result = []\n",
    "# for x in output:\n",
    "#     result.append(np.argmax(x))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "united-offering",
   "metadata": {},
   "outputs": [],
   "source": [
    "# not_pitching = [x for x in result if x == 0]\n",
    "# pitching = [x for x in result if x == 1]\n",
    "\n",
    "# print(len(not_pitching))\n",
    "# print(len(pitching))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "sunrise-noise",
   "metadata": {},
   "outputs": [],
   "source": [
    "# outputs[213]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "informational-cheat",
   "metadata": {},
   "outputs": [],
   "source": [
    "# import shutil\n",
    "# from glob import glob\n",
    "# from sklearn.model_selection import train_test_split"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "historic-remove",
   "metadata": {},
   "outputs": [],
   "source": [
    "# src = '../0310_dataset'\n",
    "# dst = '../0310_dataset/train'"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "sticky-singing",
   "metadata": {},
   "source": [
    "### Train 폴더에 데이터 전부 옮기기"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "vanilla-royalty",
   "metadata": {},
   "outputs": [],
   "source": [
    "# dir_list = os.listdir(src)\n",
    "# dir_list.remove('train')\n",
    "# dir_list.remove('validation')\n",
    "# print(dir_list)\n",
    "# cnt = 0\n",
    "# for x in dir_list:\n",
    "#     for y in os.listdir(f'{src}/{x}'):\n",
    "#         if y.endswith('.jpg'):\n",
    "#             print(y)\n",
    "#             shutil.move(f'{src}/{x}/{y}', f'{dst}/{str(cnt).zfill(5)}.jpg')\n",
    "#             shutil.move(f'{src}/{x}/{y.replace(\".jpg\", \".txt\")}', f'{dst}/{str(cnt).zfill(5)}.txt')\n",
    "#             cnt += 1\n",
    "# print(cnt)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "precious-linux",
   "metadata": {},
   "outputs": [],
   "source": [
    "# data = os.listdir(f'{src}/train')\n",
    "# img_list = []\n",
    "\n",
    "# for x in data:\n",
    "#     if x.endswith('.jpg'):\n",
    "#         img_list.append(x)\n",
    "\n",
    "\n",
    "# print(len(img_list))\n",
    "# # print(img_list)\n",
    "\n",
    "# train_img_list, val_img_list = train_test_split(img_list, test_size=0.2, random_state=42)\n",
    "\n",
    "# print(len(train_img_list), len(val_img_list))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "tested-western",
   "metadata": {},
   "outputs": [],
   "source": [
    "# for x in val_img_list:\n",
    "#     shutil.move(f'{src}/train/{x}', f'{src}/validation/{x}')"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "tensor",
   "language": "python",
   "name": "tensor"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.8.0"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}