diff --git a/examples/ai/face_similarity.ipynb b/examples/ai/face_similarity.ipynb index 5938bea4a4d..d849a67fbcc 100644 --- a/examples/ai/face_similarity.ipynb +++ b/examples/ai/face_similarity.ipynb @@ -42,7 +42,7 @@ }, { "cell_type": "code", - "execution_count": 18, + "execution_count": 2, "id": "dc6b0bc2-b95f-4190-bf77-fa2dc57fc247", "metadata": {}, "outputs": [ @@ -50,6 +50,8 @@ "name": "stderr", "output_type": "stream", "text": [ + "/Users/oliverrice/Documents/supabase/supabase/examples/ai/venv/lib/python3.10/site-packages/tqdm/auto.py:21: TqdmWarning: IProgress not found. Please update jupyter and ipywidgets. See https://ipywidgets.readthedocs.io/en/stable/user_install.html\n", + " from .autonotebook import tqdm as notebook_tqdm\n", "Found cached dataset parquet (/Users/oliverrice/.cache/huggingface/datasets/ashraq___parquet/default-f6987358ed4d9f01/0.0.0/2a3b91fbd88a2c90d1dbbb32b460cf621d31bd5b05b934492fdef7d8d6f236ec)\n" ] }, @@ -62,7 +64,7 @@ "})" ] }, - "execution_count": 18, + "execution_count": 2, "metadata": {}, "output_type": "execute_result" } @@ -217,6 +219,7 @@ "# create vector store client\n", "vx = vecs.create_client(DB_CONNECTION)\n", "\n", + "vx.delete_collection('faces')\n", "# create a PostgreSQL/pgvector table named \"faces\" to contain the face embeddings\n", "faces = vx.create_collection(name=\"faces\", dimension=128)" ] @@ -273,7 +276,11 @@ " # is produced. We'll skip those cases\n", " if len(embeddings) == 1:\n", " embedding = embeddings[0]\n", - " records.append((f\"{ix}\", embedding, {}))" + " records.append((\n", + " f\"{ix}\",\n", + " embedding,\n", + " {k: v for k, v in person.items() if k != 'image'}\n", + " ))" ] }, { @@ -327,330 +334,38 @@ }, { "cell_type": "code", - "execution_count": 13, - "id": "b6596cd4-836a-4f1d-b21a-49a6e11b4fed", + "execution_count": 10, + "id": "8673ff6e-31e5-4344-990d-bf3c84be5824", "metadata": {}, - "outputs": [ - { - "name": "stdout", - "output_type": "stream", - "text": [ - "Look Alikes\n" - ] - }, - { - "data": { - "text/html": [ - "\n", - " \n", - "
\n", - " \n", - " \n", - " \n", - "
\n", - " " - ], - "text/plain": [ - "" - ] - }, - "metadata": {}, - "output_type": "display_data" - }, - { - "data": { - "text/html": [ - "\n", - " \n", - "
\n", - "
\n", - "
\n", - "

David Hasselhoff

\n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - "
\n", - "
\n", - " \n", - "
\n", - "
\n", - "

Guillermo García Cantú

\n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - "
\n", - "
\n", - " \n", - "
\n", - "
\n", - "

Michael Madsen

\n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - "
\n", - "
\n", - "
" - ], - "text/plain": [ - "" - ] - }, - "metadata": {}, - "output_type": "display_data" - } - ], + "outputs": [], "source": [ - "import ipyplot\n", + "from IPython.core.display import HTML\n", + "from typing import Dict, Any\n", "\n", - "# Grab a random face from the dataset\n", - "query_face = people[1522]['image']\n", + "def render_similar_faces(person_image: Image, faces_collection: vecs.Collection) -> HTML:\n", + " # create query face embedding\n", + " face_embedding = face_recognition.face_encodings(np.array(person_image))[0]\n", + " \n", + " # query database for similar results\n", + " result = faces_collection.query(face_embedding, limit=5, include_metadata=True)\n", + " # read metadata from the results\n", + " result_metadata = [x[1] for x in result]\n", "\n", - "# Create the face's embedding\n", - "query_embedding = face_recognition.face_encodings(np.array(query_face))[0]\n", - "\n", - "# Query the collection for the 3 most similar faces (which will include their own)\n", - "top_n = faces.query(\n", - " query_vector=query_embedding,\n", - " limit = 3\n", - ")\n", - "\n", - "# Display the result\n", - "look_alikes = [people[int(image_id)] for image_id in top_n]\n", - "images = [person['image'] for person in look_alikes]\n", - "labels = [person['name'] for person in look_alikes]\n", - "print(\"Look Alikes\")\n", - "ipyplot.plot_images(images, labels, img_width=150)" + " captioned_images = []\n", + " for metadata in result_metadata:\n", + " html = f\"\"\"\n", + "
\n", + " \n", + "
{metadata[\"name\"]}
\n", + "
\n", + " \"\"\"\n", + " captioned_images.append(html)\n", + " \n", + " return HTML(data=f\"\"\"\n", + "
\n", + " {\"\".join(captioned_images)}\n", + "
\n", + " \"\"\")" ] }, { @@ -661,1930 +376,175 @@ "## Example Results\n", "\n", "For convenience we'll create a function to search the database and try test it out on several celebrities. Since our query faces\n", - "are also in the dataset, the query face should always be first in the results." + "are also in the dataset, the query face is the first in the result output." ] }, { "cell_type": "code", - "execution_count": 14, - "id": "f53b4b4f-3a55-4a15-9fc8-02fbe10d6b54", - "metadata": {}, - "outputs": [], - "source": [ - "def display_search_results(query_face: Image, faces_collection: vecs.Collection):\n", - " print(\"Query Face\")\n", - " ipyplot.plot_images([query_face], [\"Query Face\"], img_width=150)\n", - "\n", - " query_embedding = face_recognition.face_encodings(np.array(query_face))[0]\n", - " \n", - " top_n = faces_collection.query(\n", - " query_vector=query_embedding,\n", - " limit = 6\n", - " )\n", - " look_alikes = [people[int(image_id)] for image_id in top_n]\n", - " images = [person['image'] for person in look_alikes]\n", - " labels = [person['name'] for person in look_alikes]\n", - " print(\"Look Alikes\")\n", - " ipyplot.plot_images(images, labels, img_width=150)" - ] - }, - { - "cell_type": "code", - "execution_count": 15, + "execution_count": 11, "id": "2e5f05e8-5fc3-42e8-9a3b-4b3c7ce0dc4a", "metadata": {}, "outputs": [ - { - "name": "stdout", - "output_type": "stream", - "text": [ - "Query Face\n" - ] - }, { "data": { "text/html": [ "\n", - " \n", - "
\n", - " \n", - " \n", - " \n", - "
\n", " " ], "text/plain": [ "" ] }, + "execution_count": 11, "metadata": {}, - "output_type": "display_data" - }, - { - "data": { - "text/html": [ - "\n", - " \n", - "
\n", - "
\n", - "
\n", - "

Query Face

\n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - "
\n", - "
\n", - "
" - ], - "text/plain": [ - "" - ] - }, - "metadata": {}, - "output_type": "display_data" - }, - { - "name": "stdout", - "output_type": "stream", - "text": [ - "Look Alikes\n" - ] - }, - { - "data": { - "text/html": [ - "\n", - " \n", - "
\n", - " \n", - " \n", - " \n", - "
\n", - " " - ], - "text/plain": [ - "" - ] - }, - "metadata": {}, - "output_type": "display_data" - }, - { - "data": { - "text/html": [ - "\n", - " \n", - "
\n", - "
\n", - "
\n", - "

Lucy Liu

\n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - "
\n", - "
\n", - " \n", - "
\n", - "
\n", - "

Tamlyn Tomita

\n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - "
\n", - "
\n", - " \n", - "
\n", - "
\n", - "

Constance Wu

\n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - "
\n", - "
\n", - " \n", - "
\n", - "
\n", - "

Lauren Tom

\n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - "
\n", - "
\n", - " \n", - "
\n", - "
\n", - "

Arden Cho

\n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - "
\n", - "
\n", - " \n", - "
\n", - "
\n", - "

Olivia Cheng

\n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - "
\n", - "
\n", - "
" - ], - "text/plain": [ - "" - ] - }, - "metadata": {}, - "output_type": "display_data" + "output_type": "execute_result" } ], "source": [ - "display_search_results(query_face=people[1014]['image'], faces_collection=faces)" + "render_similar_faces(\n", + " person_image=people[1014]['image'],\n", + " faces_collection=faces\n", + ")" ] }, { "cell_type": "code", - "execution_count": 16, + "execution_count": 12, "id": "1aa257bf-d1c9-4364-a6ef-46740916ace1", "metadata": {}, "outputs": [ - { - "name": "stdout", - "output_type": "stream", - "text": [ - "Query Face\n" - ] - }, { "data": { "text/html": [ "\n", - " \n", - "
\n", - " \n", - " \n", - " \n", - "
\n", " " ], "text/plain": [ "" ] }, + "execution_count": 12, "metadata": {}, - "output_type": "display_data" - }, - { - "data": { - "text/html": [ - "\n", - " \n", - "
\n", - "
\n", - "
\n", - "

Query Face

\n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - "
\n", - "
\n", - "
" - ], - "text/plain": [ - "" - ] - }, - "metadata": {}, - "output_type": "display_data" - }, - { - "name": "stdout", - "output_type": "stream", - "text": [ - "Look Alikes\n" - ] - }, - { - "data": { - "text/html": [ - "\n", - " \n", - "
\n", - " \n", - " \n", - " \n", - "
\n", - " " - ], - "text/plain": [ - "" - ] - }, - "metadata": {}, - "output_type": "display_data" - }, - { - "data": { - "text/html": [ - "\n", - " \n", - "
\n", - "
\n", - "
\n", - "

Morgan Freeman

\n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - "
\n", - "
\n", - " \n", - "
\n", - "
\n", - "

Frankie Faison

\n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - "
\n", - "
\n", - " \n", - "
\n", - "
\n", - "

Dennis Haysbert

\n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - "
\n", - "
\n", - " \n", - "
\n", - "
\n", - "

Alvin Sanders

\n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - "
\n", - "
\n", - " \n", - "
\n", - "
\n", - "

Wendell Pierce

\n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - "
\n", - "
\n", - " \n", - "
\n", - "
\n", - "

Blu Mankuma

\n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - "
\n", - "
\n", - "
" - ], - "text/plain": [ - "" - ] - }, - "metadata": {}, - "output_type": "display_data" + "output_type": "execute_result" } ], "source": [ - "display_search_results(query_face=people[15]['image'], faces_collection=faces)" + "render_similar_faces(\n", + " person_image=people[15]['image'],\n", + " faces_collection=faces\n", + ")" ] }, { "cell_type": "code", - "execution_count": 17, + "execution_count": 13, "id": "031c13f7-c830-45e0-b36e-7c1d71eca383", "metadata": {}, "outputs": [ - { - "name": "stdout", - "output_type": "stream", - "text": [ - "Query Face\n" - ] - }, { "data": { "text/html": [ "\n", - " \n", - "
\n", - " \n", - " \n", - " \n", - "
\n", " " ], "text/plain": [ "" ] }, + "execution_count": 13, "metadata": {}, - "output_type": "display_data" - }, - { - "data": { - "text/html": [ - "\n", - " \n", - "
\n", - "
\n", - "
\n", - "

Query Face

\n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - "
\n", - "
\n", - "
" - ], - "text/plain": [ - "" - ] - }, - "metadata": {}, - "output_type": "display_data" - }, - { - "name": "stdout", - "output_type": "stream", - "text": [ - "Look Alikes\n" - ] - }, - { - "data": { - "text/html": [ - "\n", - " \n", - "
\n", - " \n", - " \n", - " \n", - "
\n", - " " - ], - "text/plain": [ - "" - ] - }, - "metadata": {}, - "output_type": "display_data" - }, - { - "data": { - "text/html": [ - "\n", - " \n", - "
\n", - "
\n", - "
\n", - "

Sandra Bullock

\n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - "
\n", - "
\n", - " \n", - "
\n", - "
\n", - "

Rebecca Ferguson

\n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - "
\n", - "
\n", - " \n", - "
\n", - "
\n", - "

Eugenie Bondurant

\n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - "
\n", - "
\n", - " \n", - "
\n", - "
\n", - "

Joyce Hyser

\n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - "
\n", - "
\n", - " \n", - "
\n", - "
\n", - "

Zuhal Olcay

\n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - "
\n", - "
\n", - " \n", - "
\n", - "
\n", - "

Marcia Gay Harden

\n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - "
\n", - "
\n", - "
" - ], - "text/plain": [ - "" - ] - }, - "metadata": {}, - "output_type": "display_data" + "output_type": "execute_result" } ], "source": [ - "display_search_results(query_face=people[188]['image'], faces_collection=faces)" + "render_similar_faces(\n", + " person_image=people[188]['image'],\n", + " faces_collection=faces\n", + ")" ] } ],