MP.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "MP.ipynb",
      "provenance": [],
      "collapsed_sections": []
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    }
  },
  "cells": [
    {
      "cell_type": "code",
      "metadata": {
        "id": "BD05yayN438n",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        ""
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "v6O64CQN8MsK",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "import os\n",
        "count=0;\n",
        "import cv2"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "v7izN6Um5Dwg",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "for filename in os.listdir('/content/fight-detection-surv-dataset/fight'):\n",
        "  #print(filename)\n",
        "  vidObj = cv2.VideoCapture('/content/fight-detection-surv-dataset/fight/'+filename) \n",
        "  while True: \n",
        "        success, image = vidObj.read() \n",
        "        #print(success)\n",
        "        # Saves the frames with frame-count \n",
        "        if success:\n",
        "          cv2.resize(image,(240,240));\n",
        "          cv2.imwrite(\"/content/fight-detection-surv-dataset/Dataset/fight1/frame%d.jpg\" % count, image) \n",
        "          count += 1\n",
        "        else:\n",
        "          break\n",
        "  vidObj.release() \n",
        "  cv2.destroyAllWindows() "
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "IiElyjBTZqal",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "count=0"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "P8T3yQ-b8JZZ",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "for filename in os.listdir('/content/fight-detection-surv-dataset/noFight'):\n",
        "  #print(filename)\n",
        "  vidObj = cv2.VideoCapture('/content/fight-detection-surv-dataset/noFight/'+filename) \n",
        "  while True: \n",
        "        success, image = vidObj.read() \n",
        "        #print(success)\n",
        "        # Saves the frames with frame-count \n",
        "        if success:\n",
        "          cv2.resize(image,(240,240));\n",
        "          cv2.imwrite(\"/content/fight-detection-surv-dataset/Dataset/nonfight1/frame%d.jpg\" % count, image) \n",
        "          count += 1\n",
        "        else:\n",
        "          break\n",
        "  vidObj.release() \n",
        "  cv2.destroyAllWindows() "
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "HSyi1SZvFRww",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        },
        "outputId": "5372bfec-c632-4b54-fe22-e442aa5d898f"
      },
      "source": [
        "from keras.applications.inception_v3 import InceptionV3\n",
        "from keras.optimizers import SGD\n",
        "from keras.preprocessing.image import ImageDataGenerator\n",
        "from keras.models import Model\n",
        "from keras.layers import Dense, GlobalAveragePooling2D\n",
        "from keras.callbacks import ModelCheckpoint, TensorBoard, EarlyStopping\n",
        "#from data import DataSet\n",
        "import os.path"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Using TensorFlow backend.\n"
          ],
          "name": "stderr"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "C0p63fXvXHHq",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "def get_model(weights='imagenet'):\n",
        "    # create the base pre-trained model\n",
        "    base_model = InceptionV3(weights=weights, include_top=False)\n",
        "\n",
        "    # add a global spatial average pooling layer\n",
        "    x = base_model.output\n",
        "    x = GlobalAveragePooling2D()(x)\n",
        "    # let's add a fully-connected layer\n",
        "    x = Dense(1024, activation='relu')(x)\n",
        "    # and a logistic layer\n",
        "    predictions = Dense(2, activation='softmax')(x)\n",
        "\n",
        "    # this is the model we will train\n",
        "    model = Model(inputs=base_model.input, outputs=predictions)\n",
        "    return model"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "-vhPShG2XVDA",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "def get_generators():\n",
        "    train_datagen = ImageDataGenerator(\n",
        "        rescale=1./255,\n",
        "        shear_range=0.2,\n",
        "        horizontal_flip=True,\n",
        "        rotation_range=10.,\n",
        "        width_shift_range=0.2,\n",
        "        height_shift_range=0.2)\n",
        "\n",
        "    test_datagen = ImageDataGenerator(rescale=1./255)\n",
        "\n",
        "    train_generator = train_datagen.flow_from_directory(\n",
        "        '/content/fight-detection-surv-dataset/Dataset',\n",
        "        target_size=(299, 299),\n",
        "        batch_size=32,\n",
        "        class_mode='categorical')\n",
        "\n",
        "    return train_generator,validation_generator"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "7QP9emMuX2WK",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "model = get_model()\n",
        "generators = get_generators()"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "zcQCLfJPJtc2",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "checkpointer = ModelCheckpoint(\n",
        "    filepath=os.path.join('data', 'checkpoints', 'inception.{epoch:03d}.hdf5'),\n",
        "    verbose=1,\n",
        "    save_best_only=True)\n",
        "\n",
        "# Helper: Stop when we stop learning.\n",
        "early_stopper = EarlyStopping(patience=10)\n",
        "\n",
        "# Helper: TensorBoard\n",
        "tensorboard = TensorBoard(log_dir=os.path.join('data', 'logs'))"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "rQ2CkkyGYBFT",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "def train_model(model, nb_epoch, generators, callbacks=[]):\n",
        "    train_generator =generators\n",
        "    model.fit_generator(\n",
        "        train_generator,\n",
        "        steps_per_epoch=100,\n",
        "        epochs=nb_epoch,\n",
        "        callbacks=callbacks)\n",
        "    return model"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "InntPv3TbMRt",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "def freeze_all_but_top(model):\n",
        "    \"\"\"Used to train just the top layers of the model.\"\"\"\n",
        "    # first: train only the top layers (which were randomly initialized)\n",
        "    # i.e. freeze all convolutional InceptionV3 layers\n",
        "    for layer in model.layers[:-2]:\n",
        "        layer.trainable = False\n",
        "\n",
        "    # compile the model (should be done *after* setting layers to non-trainable)\n",
        "    model.compile(optimizer='rmsprop', loss='categorical_crossentropy', metrics=['accuracy'])\n",
        "\n",
        "    return model\n",
        "\n",
        "def freeze_all_but_mid_and_top(model):\n",
        "    \"\"\"After we fine-tune the dense layers, train deeper.\"\"\"\n",
        "    # we chose to train the top 2 inception blocks, i.e. we will freeze\n",
        "    # the first 172 layers and unfreeze the rest:\n",
        "    for layer in model.layers[:172]:\n",
        "        layer.trainable = False\n",
        "    for layer in model.layers[172:]:\n",
        "        layer.trainable = True\n",
        "\n",
        "    # we need to recompile the model for these modifications to take effect\n",
        "    # we use SGD with a low learning rate\n",
        "    model.compile(\n",
        "        optimizer=SGD(lr=0.0001, momentum=0.9),\n",
        "        loss='categorical_crossentropy',\n",
        "        metrics=['accuracy', 'top_k_categorical_accuracy'])\n",
        "\n",
        "    return model"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "uKGzI1iEbUH5",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 353
        },
        "outputId": "2793b919-09ea-4e15-ace6-d7b545baf8c4"
      },
      "source": [
        "model = freeze_all_but_top(model)\n",
        "model = train_model(model, 10, generators)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Epoch 1/10\n",
            "100/100 [==============================] - 841s 8s/step - loss: 1.0963 - accuracy: 0.5888\n",
            "Epoch 2/10\n",
            "100/100 [==============================] - 838s 8s/step - loss: 0.6456 - accuracy: 0.6769\n",
            "Epoch 3/10\n",
            "100/100 [==============================] - 850s 9s/step - loss: 0.6277 - accuracy: 0.6809\n",
            "Epoch 4/10\n",
            "100/100 [==============================] - 848s 8s/step - loss: 0.5567 - accuracy: 0.7125\n",
            "Epoch 5/10\n",
            "100/100 [==============================] - 837s 8s/step - loss: 0.5416 - accuracy: 0.7250\n",
            "Epoch 6/10\n",
            "100/100 [==============================] - 838s 8s/step - loss: 0.5187 - accuracy: 0.7300\n",
            "Epoch 7/10\n",
            "100/100 [==============================] - 837s 8s/step - loss: 0.4804 - accuracy: 0.7609\n",
            "Epoch 8/10\n",
            "100/100 [==============================] - 913s 9s/step - loss: 0.4592 - accuracy: 0.7788\n",
            "Epoch 9/10\n",
            "100/100 [==============================] - 920s 9s/step - loss: 0.4366 - accuracy: 0.7816\n",
            "Epoch 10/10\n",
            "100/100 [==============================] - 914s 9s/step - loss: 0.4047 - accuracy: 0.8025\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "DhBWjvigbbNt",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 67
        },
        "outputId": "b65d69f1-6fe9-4efc-a9f1-bac220aee87d"
      },
      "source": [
        "  for file in os.listdir('/content/fight-detection-surv-dataset/Dataset'):\n",
        "  print('/content/fight-detection-surv-dataset/Dataset/'+file)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "/content/fight-detection-surv-dataset/Dataset/.ipynb_checkpoints\n",
            "/content/fight-detection-surv-dataset/Dataset/nonfight1\n",
            "/content/fight-detection-surv-dataset/Dataset/fight1\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Mk6WFoqrb0Px",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        },
        "outputId": "8249691d-e9e0-46c6-c204-61f7e4e122aa"
      },
      "source": [
        "import shutil \n",
        "\n",
        "location=\"/content/fight-detection-surv-dataset/Dataset\"\n",
        "path = os.path.join(location, '.ipynb_checkpoints') \n",
        "print(path)\n",
        "# removing directory \n",
        "shutil.rmtree(path) "
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "/content/fight-detection-surv-dataset/Dataset/.ipynb_checkpoints\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "cvqo17FhcRJD",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 50
        },
        "outputId": "c7db1c04-fa61-4308-86f0-45f7a6c94556"
      },
      "source": [
        "for file in os.listdir('/content/fight-detection-surv-  dataset/Dataset'):\n",
        "  print('/content/fight-detection-surv-dataset/Dataset/'+file)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "/content/fight-detection-surv-dataset/Dataset/nonfight1\n",
            "/content/fight-detection-surv-dataset/Dataset/fight1\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "2_GBvp-9dUzj",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 440
        },
        "outputId": "debf31bd-7e54-483d-dc49-62e193a84e7a"
      },
      "source": [
        "model = freeze_all_but_mid_and_top(model)\n",
        "model = train_model(model, 10, generators,\n",
        "                        [checkpointer, early_stopper, tensorboard])"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Epoch 1/10\n",
            "100/100 [==============================] - 1593s 16s/step - loss: 0.2855 - accuracy: 0.8750 - top_k_categorical_accuracy: 1.0000\n",
            "Epoch 2/10\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "/usr/local/lib/python3.6/dist-packages/keras/callbacks/callbacks.py:707: RuntimeWarning: Can save best model only with val_loss available, skipping.\n",
            "  'skipping.' % (self.monitor), RuntimeWarning)\n",
            "/usr/local/lib/python3.6/dist-packages/keras/callbacks/callbacks.py:846: RuntimeWarning: Early stopping conditioned on metric `val_loss` which is not available. Available metrics are: loss,accuracy,top_k_categorical_accuracy\n",
            "  (self.monitor, ','.join(list(logs.keys()))), RuntimeWarning\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "100/100 [==============================] - 1578s 16s/step - loss: 0.2671 - accuracy: 0.8866 - top_k_categorical_accuracy: 1.0000\n",
            "Epoch 3/10\n",
            "100/100 [==============================] - 1585s 16s/step - loss: 0.2556 - accuracy: 0.8934 - top_k_categorical_accuracy: 1.0000\n",
            "Epoch 4/10\n",
            "100/100 [==============================] - 1612s 16s/step - loss: 0.2357 - accuracy: 0.9062 - top_k_categorical_accuracy: 1.0000\n",
            "Epoch 5/10\n",
            "100/100 [==============================] - 1603s 16s/step - loss: 0.2208 - accuracy: 0.9084 - top_k_categorical_accuracy: 1.0000\n",
            "Epoch 6/10\n",
            "100/100 [==============================] - 1602s 16s/step - loss: 0.2049 - accuracy: 0.9178 - top_k_categorical_accuracy: 1.0000\n",
            "Epoch 7/10\n",
            "100/100 [==============================] - 1583s 16s/step - loss: 0.1875 - accuracy: 0.9291 - top_k_categorical_accuracy: 1.0000\n",
            "Epoch 8/10\n",
            "100/100 [==============================] - 1604s 16s/step - loss: 0.1787 - accuracy: 0.9369 - top_k_categorical_accuracy: 1.0000\n",
            "Epoch 9/10\n",
            "100/100 [==============================] - 1598s 16s/step - loss: 0.1743 - accuracy: 0.9309 - top_k_categorical_accuracy: 1.0000\n",
            "Epoch 10/10\n",
            "100/100 [==============================] - 1602s 16s/step - loss: 0.1657 - accuracy: 0.9337 - top_k_categorical_accuracy: 1.0000\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "CQaKy8TOzZRR",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 282
        },
        "outputId": "37844307-09de-4513-e2cc-021406916cf3"
      },
      "source": [
        "import matplotlib.pyplot as plt\n",
        "plt.plot(model.history.history['loss'])"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "[<matplotlib.lines.Line2D at 0x7faf9f3188d0>]"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 48
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "I-ZfESvs-4SQ",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 282
        },
        "outputId": "d9c60825-7c64-46f9-9f2a-a30dc365384f"
      },
      "source": [
        "  plt.plot(model.history.history['accuracy'])"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "[<matplotlib.lines.Line2D at 0x7faf9ee5edd8>]"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 49
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "KGGLs4UvwP5h",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "\n",
        "#for filename in os.listdir('/content/fight-detection-surv-dataset/noFight'):\n",
        "  #print(filename)\n",
        "vidObj = cv2.VideoCapture('/content/test/fi008.mp4') \n",
        "while True: \n",
        "        success, image = vidObj.read() \n",
        "        #print(success)\n",
        "        # Saves the frames with frame-count \n",
        "        if success:\n",
        "          cv2.resize(image,(299,299));\n",
        "          cv2.imwrite(\"/content/test/test_frame/frame%d.jpg\" % count, image) \n",
        "          count += 1\n",
        "        else:\n",
        "          break\n",
        "vidObj.release() \n",
        "cv2.destroyAllWindows() "
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "8U7mfriryXZw",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "import numpy as np"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Z6NYVFg5weSa",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        },
        "outputId": "029872d7-27dd-4c65-ae4f-cb3e1942c961"
      },
      "source": [
        "for image1 in os.listdir('/content/test/test_frame'):\n",
        "  image = cv2.imread(os.path.join('/content/test/test_frame',image1))\n",
        "  if image is not None:\n",
        "      #x = image.img_to_array(img)\n",
        "      x = np.expand_dims(image, axis=0)\n",
        "\n",
        "      images = np.vstack([x])\n",
        "      classes = model.predict(images, batch_size=10)\n",
        "      print(classes[0])\n",
        "      if classes[0][0]>0.5:\n",
        "          print(\"Fighting recorded\")\n",
        "      else:\n",
        "          print(\"No fight\")\n",
        "            \n",
        "  "
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "[1.000000e+00 4.653767e-24]\n",
            "Fighting recorded\n",
            "[1.0000000e+00 1.5264563e-26]\n",
            "Fighting recorded\n",
            "[1.0000000e+00 4.0827327e-27]\n",
            "Fighting recorded\n",
            "[1.0000000e+00 6.3054104e-22]\n",
            "Fighting recorded\n",
            "[1.0000000e+00 1.5512298e-24]\n",
            "Fighting recorded\n",
            "[1.0000000e+00 1.3220109e-26]\n",
            "Fighting recorded\n",
            "[1.00000000e+00 1.29726635e-27]\n",
            "Fighting recorded\n",
            "[1.000000e+00 8.692969e-21]\n",
            "Fighting recorded\n",
            "[1.0000000e+00 1.7417293e-22]\n",
            "Fighting recorded\n",
            "[1.000000e+00 4.099348e-22]\n",
            "Fighting recorded\n",
            "[1.0000000e+00 1.0421657e-24]\n",
            "Fighting recorded\n",
            "[1.0000000e+00 1.2532053e-22]\n",
            "Fighting recorded\n",
            "[1.000000e+00 6.817247e-24]\n",
            "Fighting recorded\n",
            "[1.0000000e+00 8.1557945e-29]\n",
            "Fighting recorded\n",
            "[1.000000e+00 2.352124e-25]\n",
            "Fighting recorded\n",
            "[1.0000000e+00 1.0744644e-24]\n",
            "Fighting recorded\n",
            "[1.0000000e+00 1.3421037e-24]\n",
            "Fighting recorded\n",
            "[1.000000e+00 3.429441e-26]\n",
            "Fighting recorded\n",
            "[1.0000000e+00 1.3730086e-24]\n",
            "Fighting recorded\n",
            "[1.0000000e+00 1.7146443e-26]\n",
            "Fighting recorded\n",
            "[1.0000000e+00 1.7942898e-24]\n",
            "Fighting recorded\n",
            "[1.0000000e+00 1.7555631e-23]\n",
            "Fighting recorded\n",
            "[1.000000e+00 1.166636e-24]\n",
            "Fighting recorded\n",
            "[1.000000e+00 4.406321e-22]\n",
            "Fighting recorded\n",
            "[1.0000000e+00 1.8285005e-23]\n",
            "Fighting recorded\n",
            "[1.0000000e+00 1.4422018e-26]\n",
            "Fighting recorded\n",
            "[1.0000000e+00 2.6580877e-24]\n",
            "Fighting recorded\n",
            "[1.0000000e+00 1.4557987e-27]\n",
            "Fighting recorded\n",
            "[1.0000000e+00 4.6185283e-28]\n",
            "Fighting recorded\n",
            "[1.0000000e+00 1.4948252e-20]\n",
            "Fighting recorded\n",
            "[1.000000e+00 4.089566e-24]\n",
            "Fighting recorded\n",
            "[1.0000000e+00 1.4198733e-21]\n",
            "Fighting recorded\n",
            "[1.000000e+00 7.884168e-24]\n",
            "Fighting recorded\n",
            "[1.0000000e+00 2.0409316e-21]\n",
            "Fighting recorded\n",
            "[1.0000000e+00 3.0432226e-26]\n",
            "Fighting recorded\n",
            "[1.0000000e+00 1.2902608e-23]\n",
            "Fighting recorded\n",
            "[1.0000000e+00 3.7439924e-22]\n",
            "Fighting recorded\n",
            "[1.000000e+00 4.419447e-24]\n",
            "Fighting recorded\n",
            "[1.0000000e+00 1.5358986e-22]\n",
            "Fighting recorded\n",
            "[1.000000e+00 1.702611e-20]\n",
            "Fighting recorded\n",
            "[1.0000000e+00 2.1523353e-27]\n",
            "Fighting recorded\n",
            "[1.0000000e+00 1.5396571e-24]\n",
            "Fighting recorded\n",
            "[1.0000000e+00 2.2298058e-22]\n",
            "Fighting recorded\n",
            "[1.00000e+00 5.37769e-22]\n",
            "Fighting recorded\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "emupGgLRxjf_",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        ""
      ],
      "execution_count": null,
      "outputs": []
    }
  ]
}