How to use the classification endpoint in weaviate

demo-weaviate-classification.ipynb

{
 "cells": [
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Introduction"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "This notebook shows how to upload a dataset in weaviate to be able to use the [classification endpoint](https://weaviate.io/developers/weaviate/api/rest/classification)."
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Imports"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import os\n",
    "import weaviate"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Setup"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Connect to weaviate and clear the database:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "username = os.getenv(\"WEAVIATE_USERNAME\")\n",
    "password = os.getenv(\"WEAVIATE_PASSWORD\")\n",
    "\n",
    "url = \"https://jinaai.weaviate.network/\"\n",
    "\n",
    "secret = weaviate.AuthClientPassword(username=username, password=password)\n",
    "\n",
    "client = weaviate.Client(url, auth_client_secret=secret)\n",
    "\n",
    "client.schema.delete_all()"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Dataset"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Make up some dummy dataset."
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "First, we create some labels in weaviate:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "labels = [\"chair\", \"standing desk\", \"tree\"]\n",
    "\n",
    "\n",
    "uuids = [\n",
    "    client.data_object.create(data_object={\"name\": label}, class_name=\"Label\")\n",
    "    for label in labels\n",
    "]\n",
    "\n",
    "label2uuid = {label: uuid for label, uuid in zip(labels, uuids)}"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Next, create a Mesh class with a cross reference to the Label class:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "client.schema.create_class(\n",
    "    {\n",
    "        \"class\": \"Mesh\",\n",
    "        \"properties\": [\n",
    "            {\"name\": \"name\", \"dataType\": [\"string\"]},\n",
    "            {\"name\": \"label\", \"dataType\": [\"Label\"]},\n",
    "            {\"name\": \"file\", \"dataType\": [\"string\"]},\n",
    "            {\"name\": \"description\", \"dataType\": [\"text\"]},\n",
    "        ],\n",
    "    }\n",
    ")"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Upload some meshes:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "mesh_1 = {\n",
    "    \"label\": \"chair\",\n",
    "    \"vector\": [1, 0],\n",
    "    \"file\": \"chair.png\",\n",
    "    \"description\": \"Experience unparalleled comfort with this ergonomically designed chair that conforms to your body's natural curves, providing support and reducing strain on your back\",\n",
    "}\n",
    "\n",
    "mesh_2 = {\n",
    "    \"label\": \"standing desk\",\n",
    "    \"vector\": [-1, 0],\n",
    "    \"file\": \"standing_desk.png\",\n",
    "    \"description\": \"Indulge in the ultimate symbol of luxury with a one-of-a-kind standing desk encrusted with rare jewels and precious metals, featuring an interactive holographic display and cutting-edge technology, worth billions of dollars.\",\n",
    "}\n",
    "\n",
    "meshes = [mesh_1, mesh_2]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "for mesh in meshes:\n",
    "    vector = mesh.pop(\"vector\")\n",
    "    label = mesh.pop(\"label\")\n",
    "\n",
    "    label_uuid = label2uuid[label]\n",
    "\n",
    "    mesh_uuid = client.data_object.create(\n",
    "        data_object=mesh, class_name=\"Mesh\", vector=vector\n",
    "    )\n",
    "\n",
    "    client.data_object.reference.add(\n",
    "        from_uuid=mesh_uuid,\n",
    "        from_property_name=\"label\",\n",
    "        to_uuid=label_uuid,\n",
    "        from_class_name=\"Mesh\",\n",
    "        to_class_name=\"Label\",\n",
    "    )"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Classification"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Upload a mesh without a label:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'5843a613-307d-4a50-b967-9052f7a6b014'"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "mesh_3 = {\n",
    "    \"vector\": [1, 0.1],\n",
    "    \"file\": \"sofa.png\",\n",
    "    \"description\": \"Upgrade your living space without breaking the bank with a high-quality second hand sofa that is not only affordable but also sustainable.\",\n",
    "}\n",
    "\n",
    "vector = mesh_3.pop(\"vector\")\n",
    "\n",
    "client.data_object.create(data_object=mesh_3, class_name=\"Mesh\", vector=vector)"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Make the classification request:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "classification_status = (\n",
    "    client.classification.schedule()\n",
    "    .with_type(\"zeroshot\")\n",
    "    .with_class_name(\"Mesh\")\n",
    "    .with_classify_properties([\"label\"])\n",
    "    .with_based_on_properties([\"description\"])\n",
    "    .do()\n",
    ")"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Wait for the classification to finish:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'basedOnProperties': ['description'],\n",
       " 'class': 'Mesh',\n",
       " 'classifyProperties': ['label'],\n",
       " 'id': 'd1fe6e11-684c-4249-81fb-bd119f60320c',\n",
       " 'meta': {'completed': '2023-03-07T15:24:51.447Z',\n",
       "  'count': 1,\n",
       "  'countSucceeded': 1,\n",
       "  'started': '2023-03-07T15:24:51.445Z'},\n",
       " 'status': 'completed',\n",
       " 'type': 'zeroshot'}"
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "client.classification.get(classification_status[\"id\"])"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Check the results:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'data': {'Get': {'Mesh': [{'_additional': None,\n",
       "     'file': 'standing_desk.png',\n",
       "     'label': [{'name': 'standing desk'}]},\n",
       "    {'_additional': {'classification': {'basedOn': [],\n",
       "       'classifiedFields': [],\n",
       "       'completed': '2023-03-07T15:24:51.447Z',\n",
       "       'id': 'd1fe6e11-684c-4249-81fb-bd119f60320c'}},\n",
       "     'file': 'sofa.png',\n",
       "     'label': None},\n",
       "    {'_additional': None,\n",
       "     'file': 'chair.png',\n",
       "     'label': [{'name': 'chair'}]}]}}}"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "client.query.get(\n",
    "    \"Mesh\", properties=[\"file\", \"label{... on Label{name}}\"]\n",
    ").with_additional(\n",
    "    {\"classification\": [\"basedOn\", \"classifiedFields\", \"completed\", \"id\"]}\n",
    ").do()"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Why does sofa has no label?"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Try knn classification:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "classification_status = (\n",
    "    client.classification.schedule()\n",
    "    .with_type(\"knn\")\n",
    "    .with_class_name(\"Mesh\")\n",
    "    .with_based_on_properties([\"description\"])\n",
    "    .with_classify_properties([\"label\"])\n",
    "    .with_settings({\"k\": 3})\n",
    "    .do()\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'basedOnProperties': ['description'],\n",
       " 'class': 'Mesh',\n",
       " 'classifyProperties': ['label'],\n",
       " 'id': '6f62cb92-7aaa-4545-a610-eb6ef93832da',\n",
       " 'meta': {'completed': '2023-03-07T15:24:52.225Z',\n",
       "  'count': 1,\n",
       "  'countSucceeded': 1,\n",
       "  'started': '2023-03-07T15:24:52.222Z'},\n",
       " 'settings': {'k': 3},\n",
       " 'status': 'completed',\n",
       " 'type': 'knn'}"
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "client.classification.get(classification_status[\"id\"])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'data': {'Get': {'Mesh': [{'_additional': None,\n",
       "     'file': 'standing_desk.png',\n",
       "     'label': [{'name': 'standing desk'}]},\n",
       "    {'_additional': {'classification': {'basedOn': [],\n",
       "       'classifiedFields': ['label'],\n",
       "       'completed': '2023-03-07T15:24:52.225Z',\n",
       "       'id': '6f62cb92-7aaa-4545-a610-eb6ef93832da'}},\n",
       "     'file': 'sofa.png',\n",
       "     'label': [{'name': 'chair'}]},\n",
       "    {'_additional': None,\n",
       "     'file': 'chair.png',\n",
       "     'label': [{'name': 'chair'}]}]}}}"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "client.query.get(\n",
    "    \"Mesh\", properties=[\"file\", \"label{... on Label{name}}\"]\n",
    ").with_additional(\n",
    "    {\"classification\": [\"basedOn\", \"classifiedFields\", \"completed\", \"id\"]}\n",
    ").do()"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": ".venv",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.11.2"
  },
  "orig_nbformat": 4
 },
 "nbformat": 4,
 "nbformat_minor": 2
}