Image_Simmilarity_TripletLoss.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "provenance": [],
      "gpuType": "T4",
      "mount_file_id": "1QvDNKNMeYVr46COBosk-f0rH0BotJUMq",
      "authorship_tag": "ABX9TyMXVX6ojK/tgrqeoGT0dSL4",
      "include_colab_link": true
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "language_info": {
      "name": "python"
    },
    "accelerator": "GPU"
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "view-in-github",
        "colab_type": "text"
      },
      "source": [
        "<a href=\"https://colab.research.google.com/github/MHosseinHashemi/Image_Sim/blob/main/Image_Simmilarity.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 1,
      "metadata": {
        "id": "KLGuBu_H_71L"
      },
      "outputs": [],
      "source": [
        "import tensorflow as tf\n",
        "import tensorflow_hub as hub\n",
        "import tensorflow_datasets as tfds\n",
        "import numpy as np\n",
        "import matplotlib.pyplot as plt\n",
        "import functools"
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "(train_ds, valid_ds), info = tfds.load(\"oxford_flowers102\", split=[\"train\", \"validation\"], as_supervised=True, with_info=True)"
      ],
      "metadata": {
        "id": "chsOvx8HF3um"
      },
      "execution_count": 2,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "IMG_WIDTH = IMG_HEIGHT = 256"
      ],
      "metadata": {
        "id": "J2bOnOddGGYJ"
      },
      "execution_count": 3,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "CLASSES_TO_CONSIDER = list(range(10)) # Only the first 10 classes"
      ],
      "metadata": {
        "id": "L0RApuXOGGar"
      },
      "execution_count": 4,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "# A function to Prep the images\n",
        "def preprocess_image(image, label, height, width):\n",
        "    image = tf.image.resize_with_crop_or_pad(image, target_height=height, target_width=width)\n",
        "    image = tf.cast(image, tf.float32) / 255.0\n",
        "    return image, label"
      ],
      "metadata": {
        "id": "5PfvKqYjGGdE"
      },
      "execution_count": 5,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "# A function to filter out unwanted labels ! You should remove it after test training ....\n",
        "def filter_by_classes(img, label):\n",
        "  # Only picks the selected labels\n",
        "  bools = tf.equal(label, CLASSES_TO_CONSIDER)\n",
        "  return tf.reduce_any(bools)"
      ],
      "metadata": {
        "id": "EVivHPZMGGfB"
      },
      "execution_count": 6,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "# Create a new function to prep the\n",
        "partial_preprocess_image = functools.partial(preprocess_image, height=IMG_HEIGHT, width=IMG_WIDTH)\n",
        "train_ds = train_ds.filter(filter_by_classes).map(partial_preprocess_image).cache().shuffle(buffer_size=1000)\n",
        "valid_ds = valid_ds.filter(filter_by_classes).map(partial_preprocess_image).cache().shuffle(buffer_size=1000)"
      ],
      "metadata": {
        "id": "jZbWUIWLKMRo"
      },
      "execution_count": 7,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "def get_x_y_from_ds(ds):\n",
        "    x, y = [], []\n",
        "    for img, label in ds.cache().as_numpy_iterator():\n",
        "        x.append(img)\n",
        "        y.append(label)\n",
        "\n",
        "    return np.array(x), np.array(y)"
      ],
      "metadata": {
        "id": "anwwa80rNFZ0"
      },
      "execution_count": 8,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "x_train, y_train = get_x_y_from_ds(train_ds)"
      ],
      "metadata": {
        "id": "wkr2EVrAOENB"
      },
      "execution_count": 9,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "x_test, y_test = get_x_y_from_ds(valid_ds)"
      ],
      "metadata": {
        "id": "SnZPkLLGOKAh"
      },
      "execution_count": 10,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "print(f\"number of {len(x_train)} samples\\nnumber of {len(y_train)} samples\\nnumber of {len(x_test)} samples\\nnumber of {len(y_test)} samples\")"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "tiJw2aNmOT-Q",
        "outputId": "2707170a-c8a7-4a6a-c589-afe30aad5c0e"
      },
      "execution_count": 11,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "number of 100 samples\n",
            "number of 100 samples\n",
            "number of 100 samples\n",
            "number of 100 samples\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "# Model\n",
        "MODEL_URL = \"https://tfhub.dev/google/imagenet/efficientnet_v2_imagenet1k_s/feature_vector/2\"\n",
        "\n",
        "model = tf.keras.Sequential([\n",
        "    tf.keras.layers.RandomFlip(\"horizontal_and_vertical\"),\n",
        "    tf.keras.layers.RandomRotation(0.3),\n",
        "    hub.KerasLayer(MODEL_URL, trainable=True),\n",
        "    tf.keras.layers.Dropout(0.2),\n",
        "    tf.keras.layers.Dense(256, activation=None), # No activation on final dense layer\n",
        "    tf.keras.layers.Lambda(lambda x: tf.math.l2_normalize(x, axis=1)) # L2 normalize embeddings\n",
        "])\n",
        "\n",
        "model.build([None, IMG_HEIGHT, IMG_WIDTH, 3])\n",
        "model.summary()"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "USAxUjmLOwmm",
        "outputId": "1b878830-e54b-493a-aa8d-1ee558cb1b1e"
      },
      "execution_count": 12,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Model: \"sequential\"\n",
            "_________________________________________________________________\n",
            " Layer (type)                Output Shape              Param #   \n",
            "=================================================================\n",
            " random_flip (RandomFlip)    (None, 256, 256, 3)       0         \n",
            "                                                                 \n",
            " random_rotation (RandomRota  (None, 256, 256, 3)      0         \n",
            " tion)                                                           \n",
            "                                                                 \n",
            " keras_layer (KerasLayer)    (None, 1280)              20331360  \n",
            "                                                                 \n",
            " dropout (Dropout)           (None, 1280)              0         \n",
            "                                                                 \n",
            " dense (Dense)               (None, 256)               327936    \n",
            "                                                                 \n",
            " lambda (Lambda)             (None, 256)               0         \n",
            "                                                                 \n",
            "=================================================================\n",
            "Total params: 20,659,296\n",
            "Trainable params: 20,505,424\n",
            "Non-trainable params: 153,872\n",
            "_________________________________________________________________\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "from collections import defaultdict\n",
        "import random\n",
        "\n",
        "def get_training_batch(images, labels, batch_size=128, n_examples_per_class=4):\n",
        "    examples_per_class = defaultdict(list)\n",
        "    for x, y in zip(images, labels):\n",
        "        examples_per_class[y].append(x)\n",
        "\n",
        "    while True:\n",
        "        batch_X, batch_y = [], []\n",
        "        while len(batch_X) < batch_size:\n",
        "            for cls, examples in examples_per_class.items():\n",
        "                n_sample = min(n_examples_per_class, (batch_size - len(batch_X)))\n",
        "                if n_sample == 0:\n",
        "                    break\n",
        "                samples = random.sample(examples, k=n_sample)\n",
        "                batch_X.extend(samples)\n",
        "                batch_y.extend([cls] * len(samples))\n",
        "\n",
        "        yield np.array(batch_X), np.array(batch_y)"
      ],
      "metadata": {
        "id": "WM07pLS9UqHJ"
      },
      "execution_count": 13,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "# !pip install tensorflow_addons"
      ],
      "metadata": {
        "id": "n3sRCSeXUqJo"
      },
      "execution_count": 14,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "bs = 24\n",
        "n_examples_per_class = 3\n",
        "initial_lr = 0.0001\n",
        "epochs = 80\n",
        "import tensorflow_addons as tfa\n",
        "\n",
        "model.compile(optimizer=tf.keras.optimizers.Adam(learning_rate=initial_lr),\n",
        "                         loss=tfa.losses.TripletSemiHardLoss())\n",
        "\n",
        "history = model.fit(get_training_batch(images=x_train, labels=y_train, batch_size=bs, n_examples_per_class=n_examples_per_class),\n",
        "                               epochs=epochs,\n",
        "                               steps_per_epoch=len(x_train)//bs,\n",
        "                               validation_data=(x_test, y_test),\n",
        "                               validation_batch_size=bs,\n",
        "                              )"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "0LCpHPxPUqMA",
        "outputId": "43084fcc-297b-410e-92ad-bc4f3510e81f"
      },
      "execution_count": 18,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Epoch 1/80\n",
            "4/4 [==============================] - 99s 1s/step - loss: 0.8097 - val_loss: nan\n",
            "Epoch 2/80\n",
            "4/4 [==============================] - 2s 597ms/step - loss: 0.7671 - val_loss: nan\n",
            "Epoch 3/80\n",
            "4/4 [==============================] - 2s 594ms/step - loss: 0.6984 - val_loss: nan\n",
            "Epoch 4/80\n",
            "4/4 [==============================] - 2s 595ms/step - loss: 0.6572 - val_loss: nan\n",
            "Epoch 5/80\n",
            "4/4 [==============================] - 2s 652ms/step - loss: 0.6052 - val_loss: nan\n",
            "Epoch 6/80\n",
            "4/4 [==============================] - 2s 656ms/step - loss: 0.5349 - val_loss: nan\n",
            "Epoch 7/80\n",
            "4/4 [==============================] - 2s 600ms/step - loss: 0.4920 - val_loss: nan\n",
            "Epoch 8/80\n",
            "4/4 [==============================] - 2s 648ms/step - loss: 0.4238 - val_loss: nan\n",
            "Epoch 9/80\n",
            "4/4 [==============================] - 2s 606ms/step - loss: 0.3766 - val_loss: nan\n",
            "Epoch 10/80\n",
            "4/4 [==============================] - 2s 649ms/step - loss: 0.3112 - val_loss: nan\n",
            "Epoch 11/80\n",
            "4/4 [==============================] - 2s 624ms/step - loss: 0.2747 - val_loss: nan\n",
            "Epoch 12/80\n",
            "4/4 [==============================] - 2s 661ms/step - loss: 0.2371 - val_loss: nan\n",
            "Epoch 13/80\n",
            "4/4 [==============================] - 2s 599ms/step - loss: 0.2109 - val_loss: nan\n",
            "Epoch 14/80\n",
            "4/4 [==============================] - 2s 646ms/step - loss: 0.1526 - val_loss: nan\n",
            "Epoch 15/80\n",
            "4/4 [==============================] - 2s 647ms/step - loss: 0.1673 - val_loss: nan\n",
            "Epoch 16/80\n",
            "4/4 [==============================] - 2s 596ms/step - loss: 0.1102 - val_loss: nan\n",
            "Epoch 17/80\n",
            "4/4 [==============================] - 2s 606ms/step - loss: 0.0870 - val_loss: nan\n",
            "Epoch 18/80\n",
            "4/4 [==============================] - 2s 672ms/step - loss: 0.0724 - val_loss: nan\n",
            "Epoch 19/80\n",
            "4/4 [==============================] - 2s 635ms/step - loss: 0.0749 - val_loss: nan\n",
            "Epoch 20/80\n",
            "4/4 [==============================] - 2s 646ms/step - loss: 0.0455 - val_loss: nan\n",
            "Epoch 21/80\n",
            "4/4 [==============================] - 2s 640ms/step - loss: 0.0324 - val_loss: nan\n",
            "Epoch 22/80\n",
            "4/4 [==============================] - 2s 591ms/step - loss: 0.0250 - val_loss: nan\n",
            "Epoch 23/80\n",
            "4/4 [==============================] - 2s 644ms/step - loss: 0.0215 - val_loss: nan\n",
            "Epoch 24/80\n",
            "4/4 [==============================] - 2s 623ms/step - loss: 0.0190 - val_loss: nan\n",
            "Epoch 25/80\n",
            "4/4 [==============================] - 2s 684ms/step - loss: 0.0120 - val_loss: nan\n",
            "Epoch 26/80\n",
            "4/4 [==============================] - 2s 590ms/step - loss: 0.0160 - val_loss: nan\n",
            "Epoch 27/80\n",
            "4/4 [==============================] - 2s 585ms/step - loss: 0.0115 - val_loss: nan\n",
            "Epoch 28/80\n",
            "4/4 [==============================] - 2s 637ms/step - loss: 0.0082 - val_loss: nan\n",
            "Epoch 29/80\n",
            "4/4 [==============================] - 2s 634ms/step - loss: 0.0065 - val_loss: nan\n",
            "Epoch 30/80\n",
            "4/4 [==============================] - 3s 735ms/step - loss: 0.0097 - val_loss: nan\n",
            "Epoch 31/80\n",
            "4/4 [==============================] - 2s 648ms/step - loss: 0.0059 - val_loss: nan\n",
            "Epoch 32/80\n",
            "4/4 [==============================] - 2s 635ms/step - loss: 0.0033 - val_loss: nan\n",
            "Epoch 33/80\n",
            "4/4 [==============================] - 2s 586ms/step - loss: 0.0024 - val_loss: nan\n",
            "Epoch 34/80\n",
            "4/4 [==============================] - 2s 637ms/step - loss: 0.0051 - val_loss: nan\n",
            "Epoch 35/80\n",
            "4/4 [==============================] - 2s 587ms/step - loss: 0.0017 - val_loss: nan\n",
            "Epoch 36/80\n",
            "4/4 [==============================] - 2s 645ms/step - loss: 0.0020 - val_loss: nan\n",
            "Epoch 37/80\n",
            "4/4 [==============================] - 2s 608ms/step - loss: 7.6143e-04 - val_loss: nan\n",
            "Epoch 38/80\n",
            "4/4 [==============================] - 2s 588ms/step - loss: 0.0016 - val_loss: nan\n",
            "Epoch 39/80\n",
            "4/4 [==============================] - 2s 640ms/step - loss: 0.0033 - val_loss: nan\n",
            "Epoch 40/80\n",
            "4/4 [==============================] - 2s 594ms/step - loss: 0.0018 - val_loss: nan\n",
            "Epoch 41/80\n",
            "4/4 [==============================] - 2s 642ms/step - loss: 0.0028 - val_loss: nan\n",
            "Epoch 42/80\n",
            "4/4 [==============================] - 2s 606ms/step - loss: 0.0023 - val_loss: nan\n",
            "Epoch 43/80\n",
            "4/4 [==============================] - 2s 610ms/step - loss: 0.0017 - val_loss: nan\n",
            "Epoch 44/80\n",
            "4/4 [==============================] - 2s 639ms/step - loss: 0.0010 - val_loss: nan\n",
            "Epoch 45/80\n",
            "4/4 [==============================] - 2s 633ms/step - loss: 0.0021 - val_loss: nan\n",
            "Epoch 46/80\n",
            "4/4 [==============================] - 2s 596ms/step - loss: 0.0024 - val_loss: nan\n",
            "Epoch 47/80\n",
            "4/4 [==============================] - 2s 631ms/step - loss: 1.0536e-04 - val_loss: nan\n",
            "Epoch 48/80\n",
            "4/4 [==============================] - 2s 647ms/step - loss: 3.3129e-04 - val_loss: nan\n",
            "Epoch 49/80\n",
            "4/4 [==============================] - 2s 660ms/step - loss: 0.0010 - val_loss: nan\n",
            "Epoch 50/80\n",
            "4/4 [==============================] - 2s 639ms/step - loss: 0.0015 - val_loss: nan\n",
            "Epoch 51/80\n",
            "4/4 [==============================] - 2s 589ms/step - loss: 6.4637e-04 - val_loss: nan\n",
            "Epoch 52/80\n",
            "4/4 [==============================] - 2s 628ms/step - loss: 0.0041 - val_loss: nan\n",
            "Epoch 53/80\n",
            "4/4 [==============================] - 2s 637ms/step - loss: 4.0314e-05 - val_loss: nan\n",
            "Epoch 54/80\n",
            "4/4 [==============================] - 2s 648ms/step - loss: 0.0010 - val_loss: nan\n",
            "Epoch 55/80\n",
            "4/4 [==============================] - 3s 733ms/step - loss: 9.7433e-05 - val_loss: nan\n",
            "Epoch 56/80\n",
            "4/4 [==============================] - 2s 583ms/step - loss: 0.0020 - val_loss: nan\n",
            "Epoch 57/80\n",
            "4/4 [==============================] - 2s 636ms/step - loss: 1.4398e-04 - val_loss: nan\n",
            "Epoch 58/80\n",
            "4/4 [==============================] - 2s 585ms/step - loss: 1.8479e-04 - val_loss: nan\n",
            "Epoch 59/80\n",
            "4/4 [==============================] - 2s 629ms/step - loss: 0.0017 - val_loss: nan\n",
            "Epoch 60/80\n",
            "4/4 [==============================] - 2s 641ms/step - loss: 3.6382e-04 - val_loss: nan\n",
            "Epoch 61/80\n",
            "4/4 [==============================] - 2s 642ms/step - loss: 7.0874e-04 - val_loss: nan\n",
            "Epoch 62/80\n",
            "4/4 [==============================] - 2s 582ms/step - loss: 6.5775e-04 - val_loss: nan\n",
            "Epoch 63/80\n",
            "4/4 [==============================] - 2s 633ms/step - loss: 3.5430e-04 - val_loss: nan\n",
            "Epoch 64/80\n",
            "4/4 [==============================] - 2s 629ms/step - loss: 2.3986e-04 - val_loss: nan\n",
            "Epoch 65/80\n",
            "4/4 [==============================] - 2s 589ms/step - loss: 0.0013 - val_loss: nan\n",
            "Epoch 66/80\n",
            "4/4 [==============================] - 2s 642ms/step - loss: 2.0509e-04 - val_loss: nan\n",
            "Epoch 67/80\n",
            "4/4 [==============================] - 2s 646ms/step - loss: 5.5009e-04 - val_loss: nan\n",
            "Epoch 68/80\n",
            "4/4 [==============================] - 3s 708ms/step - loss: 4.3806e-05 - val_loss: nan\n",
            "Epoch 69/80\n",
            "4/4 [==============================] - 3s 779ms/step - loss: 1.6835e-04 - val_loss: nan\n",
            "Epoch 70/80\n",
            "4/4 [==============================] - 3s 728ms/step - loss: 8.2267e-08 - val_loss: nan\n",
            "Epoch 71/80\n",
            "4/4 [==============================] - 3s 805ms/step - loss: 5.5624e-04 - val_loss: nan\n",
            "Epoch 72/80\n",
            "4/4 [==============================] - 4s 1s/step - loss: 4.7074e-04 - val_loss: nan\n",
            "Epoch 73/80\n",
            "4/4 [==============================] - 3s 721ms/step - loss: 4.0906e-04 - val_loss: nan\n",
            "Epoch 74/80\n",
            "4/4 [==============================] - 3s 751ms/step - loss: 6.0567e-04 - val_loss: nan\n",
            "Epoch 75/80\n",
            "4/4 [==============================] - 3s 784ms/step - loss: 0.0015 - val_loss: nan\n",
            "Epoch 76/80\n",
            "4/4 [==============================] - 3s 679ms/step - loss: 4.4724e-04 - val_loss: nan\n",
            "Epoch 77/80\n",
            "4/4 [==============================] - 3s 744ms/step - loss: 1.5628e-04 - val_loss: nan\n",
            "Epoch 78/80\n",
            "4/4 [==============================] - 2s 601ms/step - loss: 2.2951e-04 - val_loss: nan\n",
            "Epoch 79/80\n",
            "4/4 [==============================] - 2s 589ms/step - loss: 2.1910e-04 - val_loss: nan\n",
            "Epoch 80/80\n",
            "4/4 [==============================] - 2s 632ms/step - loss: 1.2770e-04 - val_loss: nan\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "print(\"Done\")"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "FzWTfLDqjukh",
        "outputId": "ce98fd0a-0dd4-463d-b6f4-07d83bfb16d6"
      },
      "execution_count": 19,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Done\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "# plot model performance\n",
        "\n",
        "\n",
        "loss = history.history['loss']\n",
        "val_loss = history.history['val_loss']\n",
        "epochs_range = range(1, len(history.epoch) + 1)\n",
        "\n",
        "plt.figure(figsize=(15,5))\n",
        "\n",
        "plt.plot(epochs_range, loss, label='Train Set')\n",
        "plt.plot(epochs_range, val_loss, label='Val Set')\n",
        "plt.legend(loc=\"best\")\n",
        "plt.xlabel('Epochs')\n",
        "plt.ylabel('Loss')\n",
        "plt.title('Model Loss')\n",
        "\n",
        "plt.tight_layout()\n",
        "plt.show()"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 342
        },
        "id": "4mdgV9GdyKfK",
        "outputId": "aa493067-3f0a-401b-edd7-f5c33a8f8fb1"
      },
      "execution_count": 31,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 1500x500 with 1 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {}
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "# Logs"
      ],
      "metadata": {
        "id": "a1wsabkv1kPW"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "###### - 30 Epocsh Seems to be enough\n",
        "###### - 32 batch is the maximum number\n",
        "###### - Validation Loss is not a Number !?!?!"
      ],
      "metadata": {
        "id": "fb5ahLBZ1Gfv"
      }
    },
    {
      "cell_type": "code",
      "source": [],
      "metadata": {
        "id": "FPDwmiVK1K3k"
      },
      "execution_count": null,
      "outputs": []
    }
  ]
}