batch_triplet_loss.ipynb

batch_triplet_loss.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "batch_triplet_loss.ipynb",
      "provenance": [],
      "authorship_tag": "ABX9TyMqHmAHy2SWbykAVPPGIwDz",
      "include_colab_link": true
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    }
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "view-in-github",
        "colab_type": "text"
      },
      "source": [
        "<a href=\"https://colab.research.google.com/gist/drcege/bd96768240ef3cc0a37479c5870b6e0c/batch_triplet_loss.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "W3f7ZZ_IWNGo",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "import tensorflow as tf\n",
        "import numpy as np"
      ],
      "execution_count": 1,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "3wCISCWcXD_v",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "_margin = 0.3"
      ],
      "execution_count": 2,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "OvbFXgk5XIr5",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "def euclidean_distance(x, y):\n",
        "    return tf.norm(x - y, axis=-1)\n",
        "\n",
        "dist_func = euclidean_distance"
      ],
      "execution_count": 3,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "oEj2GlRzXQZO",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "def triplet_loss(y_true, y_pred):\n",
        "    bs = tf.shape(y_pred)[0]\n",
        "    #tf.print('\\nbs', bs)\n",
        "    sketch, photo = tf.split(y_pred, 2, axis=-1)\n",
        "    \n",
        "    pd = dist_func(sketch, photo)\n",
        "    \n",
        "    ss = tf.expand_dims(sketch, 0)\n",
        "    pp = tf.expand_dims(photo, 1)\n",
        "    nd = dist_func(ss, pp)\n",
        "    \n",
        "    t_loss = pd - nd + _margin  # N * N \n",
        "    eye = tf.eye(bs)\n",
        "    tt = t_loss * (1 - eye)\n",
        "    \n",
        "    LOSS = tf.reduce_sum(tf.nn.relu(tt), axis=0) / tf.cast(bs-1, tf.float32)  # N\n",
        "    #tf.print(tf.shape(LOSS))\n",
        "    return LOSS"
      ],
      "execution_count": 4,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "XT_K5I3VXXjm",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 85
        },
        "outputId": "3c7383f1-da8e-4dcf-b028-6fe8dd10d294"
      },
      "source": [
        "A = tf.nn.l2_normalize(tf.random.uniform((3,2)), axis=-1); A"
      ],
      "execution_count": 5,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<tf.Tensor: shape=(3, 2), dtype=float32, numpy=\n",
              "array([[0.87696016, 0.48056316],\n",
              "       [0.9718337 , 0.23566774],\n",
              "       [0.6842739 , 0.72922504]], dtype=float32)>"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 5
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "jVyEiL7yXko0",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 85
        },
        "outputId": "9e1ea84b-715d-4db5-f9b8-a92850ba2460"
      },
      "source": [
        "B = tf.nn.l2_normalize(tf.random.uniform((3,2)), axis=-1); B"
      ],
      "execution_count": 6,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<tf.Tensor: shape=(3, 2), dtype=float32, numpy=\n",
              "array([[0.66403735, 0.74769926],\n",
              "       [0.66695917, 0.74509424],\n",
              "       [0.9976887 , 0.06795112]], dtype=float32)>"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 6
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "ZI9-kbs2YTdC",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 85
        },
        "outputId": "84f0816e-7b1a-4e06-c626-860ea8e6d580"
      },
      "source": [
        "C = tf.concat([A, B], axis=-1); C"
      ],
      "execution_count": 7,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<tf.Tensor: shape=(3, 4), dtype=float32, numpy=\n",
              "array([[0.87696016, 0.48056316, 0.66403735, 0.74769926],\n",
              "       [0.9718337 , 0.23566774, 0.66695917, 0.74509424],\n",
              "       [0.6842739 , 0.72922504, 0.9976887 , 0.06795112]], dtype=float32)>"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 7
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "80xAAzfkX0SM",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        },
        "outputId": "6ac5712a-ef73-42e1-f5db-e63a58d903b2"
      },
      "source": [
        "print(triplet_loss(None, C))"
      ],
      "execution_count": 8,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "tf.Tensor([0.25777826 0.510126   1.0063429 ], shape=(3,), dtype=float32)\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "qTycXF71ZdYz",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        ""
      ],
      "execution_count": 8,
      "outputs": []
    }
  ]
}