From d81b3a4bb26eb6f012ccaecd861838a0c31bb685 Mon Sep 17 00:00:00 2001
From: Tiago de Freitas Pereira <tiago.pereira@idiap.ch>
Date: Tue, 4 May 2021 18:34:00 +0200
Subject: [PATCH] 50 shades of face
---
notebooks/50-shades-of-face.ipynb | 359 ++++++++++++++++++++++++++++++
1 file changed, 359 insertions(+)
create mode 100644 notebooks/50-shades-of-face.ipynb
diff --git a/notebooks/50-shades-of-face.ipynb b/notebooks/50-shades-of-face.ipynb
new file mode 100644
index 00000000..0e987722
--- /dev/null
+++ b/notebooks/50-shades-of-face.ipynb
@@ -0,0 +1,359 @@
+{
+ "cells": [
+ {
+ "cell_type": "markdown",
+ "metadata": {},
+ "source": [
+ "# 50 Shades of face\n",
+ "\n",
+ "In this notebook we aim to evalute the impact of different face crops in FR baselines.\n",
+ "For that we are relying on the MOBIO dataset, which is not ideal, but it's short enough to run a bunch of experiments."
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": 1,
+ "metadata": {},
+ "outputs": [],
+ "source": [
+ "# Fetching resources\n",
+ "import bob.bio.base\n",
+ "from bob.bio.base.pipelines.vanilla_biometrics import execute_vanilla_biometrics\n",
+ "from bob.bio.base.pipelines.vanilla_biometrics import Distance\n",
+ "from bob.bio.base.pipelines.vanilla_biometrics import VanillaBiometricsPipeline\n",
+ "\n",
+ "from bob.bio.face.database import MobioDatabase\n",
+ "from bob.bio.face.preprocessor import FaceCrop\n",
+ "from bob.extension import rc\n",
+ "from bob.pipelines import wrap\n",
+ "import os\n",
+ "import scipy.spatial\n",
+ "import bob.measure\n",
+ "dask_client = None\n",
+ "\n",
+ "###\n",
+ "image_size = 112\n",
+ "\n",
+ "# eyes position in the vertical axis\n",
+ "# final position will be image_size/height_denominators\n",
+ "height_denominators = [4.5,4,3.5,3,2.8]\n",
+ "\n",
+ "# Eyes distance to be explored\n",
+ "eyes_distances = [30, 35, 40, 42,45,48]\n",
+ "\n",
+ "\n",
+ "\n",
+ "output_path = \"./50-shades\"\n",
+ "\n",
+ "\n",
+ "######## CHANGE YOUR FEATURE EXTRACTOR HERE\n",
+ "from bob.bio.face.embeddings.mxnet_models import ArcFaceInsightFace\n",
+ "extractor_transformer = wrap([\"sample\"],ArcFaceInsightFace())\n",
+ "\n",
+ "### CHANGE YOUR MATCHER HERE\n",
+ "algorithm = Distance(distance_function = scipy.spatial.distance.cosine,is_distance_function = True)\n",
+ "\n",
+ "\n",
+ "##### CHANGE YOUR DATABASE HERE\n",
+ "database = MobioDatabase(protocol=\"mobile0-male\")\n",
+ "sample = database.references()[0][0]\n",
+ "import matplotlib.pyplot as plt\n",
+ "import bob.io.image\n",
+ "\n",
+ "\n"
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "metadata": {},
+ "source": [
+ "## Setting up the grid\n",
+ "\n",
+ "If you want to run this on the cluster, don't forget to `SETSHELL grid` before running the cell below.\n"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": 2,
+ "metadata": {},
+ "outputs": [],
+ "source": [
+ "# Starting dask client\n",
+ "\n",
+ "\n",
+ "from dask.distributed import Client\n",
+ "from bob.pipelines.distributed.sge import SGEMultipleQueuesCluster\n",
+ "\n",
+ "cluster = SGEMultipleQueuesCluster(min_jobs=1)\n",
+ "dask_client = Client(cluster)\n"
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "metadata": {},
+ "source": [
+ "## Running different face crops\n",
+ "\n",
+ "Here we are varying the `eyes_distances` and the ration `fig_size/height_denominators`,\n",
+ "generating the transformers and plotting the outcome"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": 3,
+ "metadata": {},
+ "outputs": [
+ {
+ "data": {
+ "application/vnd.jupyter.widget-view+json": {
+ "model_id": "365e408f16e04de7ba0f709639b4ee8d",
+ "version_major": 2,
+ "version_minor": 0
+ },
+ "text/plain": [
+ "Canvas(toolbar=Toolbar(toolitems=[('Home', 'Reset original view', 'home', 'home'), ('Back', 'Back to previous …"
+ ]
+ },
+ "metadata": {},
+ "output_type": "display_data"
+ }
+ ],
+ "source": [
+ "%matplotlib widget\n",
+ "\n",
+ "annotation_type = \"eyes-center\"\n",
+ "fixed_positions = None\n",
+ "\n",
+ "\n",
+ "def get_cropers(final_size = 112,\n",
+ " height_denominators = [5,4,3,2],\n",
+ " eyes_distances = [30, 35, 40, 42,45]):\n",
+ " \n",
+ " left_eye_offset = 1.49\n",
+ "\n",
+ " transformers = []\n",
+ " for e in eyes_distances:\n",
+ " for h in height_denominators:\n",
+ "\n",
+ " right_eye_offset = (final_size*left_eye_offset)/(final_size-e*left_eye_offset)\n",
+ " \n",
+ " RIGHT_EYE_POS = (final_size / h, final_size/right_eye_offset)\n",
+ " LEFT_EYE_POS = (final_size / h, final_size/left_eye_offset)\n",
+ "\n",
+ "\n",
+ " #RIGHT_EYE_POS = (final_size / 3.44, final_size / 3.02)\n",
+ " #LEFT_EYE_POS = (final_size / 3.44, final_size / 1.49)\n",
+ " cropped_positions = {\n",
+ " \"leye\": LEFT_EYE_POS,\n",
+ " \"reye\": RIGHT_EYE_POS,\n",
+ " }\n",
+ " #print(cropped_positions)\n",
+ "\n",
+ " preprocessor_transformer = FaceCrop(cropped_image_size=(112,112),\n",
+ " cropped_positions=cropped_positions,\n",
+ " color_channel='rgb',\n",
+ " fixed_positions=fixed_positions)\n",
+ "\n",
+ " transform_extra_arguments = (None if (cropped_positions is None or fixed_positions is not None) else ((\"annotations\", \"annotations\"),))\n",
+ "\n",
+ " preprocessor_transformer = wrap([\"sample\"], preprocessor_transformer,transform_extra_arguments=transform_extra_arguments)\n",
+ " \n",
+ " transformers.append(preprocessor_transformer)\n",
+ "\n",
+ " return transformers\n",
+ "\n",
+ "def plot_faces(transformers, database, subplot_shape, fnmrs=None):\n",
+ "\n",
+ " fig, axis = plt.subplots(subplot_shape[0], subplot_shape[1])\n",
+ " offset = 0\n",
+ " for ax_h in axis:\n",
+ " for ax_w in ax_h:\n",
+ " \n",
+ " # Picking the first sample\n",
+ " sample = database.references()[0][0]\n",
+ " preprocessor_transformer = transformers[offset]\n",
+ "\n",
+ " cropped = preprocessor_transformer.transform([sample])[0]\n",
+ " cropped = bob.io.image.to_matplotlib(cropped.data).astype(\"uint8\")\n",
+ "\n",
+ " ax_w.imshow(cropped)\n",
+ " reye_y = round(preprocessor_transformer.estimator.cropped_positions[\"reye\"][0],2)\n",
+ " reye_x = round(preprocessor_transformer.estimator.cropped_positions[\"reye\"][1],2)\n",
+ " leye_y = round(preprocessor_transformer.estimator.cropped_positions[\"leye\"][0],2)\n",
+ " leye_x = round(preprocessor_transformer.estimator.cropped_positions[\"leye\"][1],2)\n",
+ " \n",
+ " if fnmrs is None:\n",
+ " title = f\"({reye_y},{reye_x}) - ({leye_y},{leye_x})\"\n",
+ " else:\n",
+ " title = f\"({reye_y},{reye_x}) - ({leye_y},{leye_x}) = {fnmrs[offset]}\"\n",
+ "\n",
+ " ax_w.set_title(f\"{title}\", fontsize=5)\n",
+ " ax_w.axis('off')\n",
+ " \n",
+ " offset +=1\n",
+ " \n",
+ "\n",
+ "subplot_shape = (int((len(eyes_distances)*len(height_denominators))/len(height_denominators)),len(height_denominators))\n",
+ "\n",
+ "transformers = get_cropers(final_size=image_size,\n",
+ " height_denominators=height_denominators,\n",
+ " eyes_distances=eyes_distances)\n",
+ "\n",
+ "plot_faces(transformers, database, subplot_shape)\n"
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "metadata": {},
+ "source": [
+ "## Run vanilla biometrics\n",
+ "\n",
+ "Here we are running Vanilla Biometrics several times and collecting the `1-FNMR@FMR=0.001` and plotting."
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": 4,
+ "metadata": {},
+ "outputs": [
+ {
+ "name": "stderr",
+ "output_type": "stream",
+ "text": [
+ "There's no data to train background model.For the rest of the execution it will be assumed that the pipeline is stateless.\n",
+ "There's no data to train background model.For the rest of the execution it will be assumed that the pipeline is stateless.\n",
+ "There's no data to train background model.For the rest of the execution it will be assumed that the pipeline is stateless.\n",
+ "There's no data to train background model.For the rest of the execution it will be assumed that the pipeline is stateless.\n",
+ "There's no data to train background model.For the rest of the execution it will be assumed that the pipeline is stateless.\n",
+ "There's no data to train background model.For the rest of the execution it will be assumed that the pipeline is stateless.\n",
+ "There's no data to train background model.For the rest of the execution it will be assumed that the pipeline is stateless.\n",
+ "There's no data to train background model.For the rest of the execution it will be assumed that the pipeline is stateless.\n",
+ "There's no data to train background model.For the rest of the execution it will be assumed that the pipeline is stateless.\n",
+ "There's no data to train background model.For the rest of the execution it will be assumed that the pipeline is stateless.\n",
+ "There's no data to train background model.For the rest of the execution it will be assumed that the pipeline is stateless.\n",
+ "There's no data to train background model.For the rest of the execution it will be assumed that the pipeline is stateless.\n",
+ "There's no data to train background model.For the rest of the execution it will be assumed that the pipeline is stateless.\n",
+ "There's no data to train background model.For the rest of the execution it will be assumed that the pipeline is stateless.\n",
+ "There's no data to train background model.For the rest of the execution it will be assumed that the pipeline is stateless.\n",
+ "There's no data to train background model.For the rest of the execution it will be assumed that the pipeline is stateless.\n",
+ "There's no data to train background model.For the rest of the execution it will be assumed that the pipeline is stateless.\n",
+ "There's no data to train background model.For the rest of the execution it will be assumed that the pipeline is stateless.\n",
+ "There's no data to train background model.For the rest of the execution it will be assumed that the pipeline is stateless.\n",
+ "There's no data to train background model.For the rest of the execution it will be assumed that the pipeline is stateless.\n",
+ "There's no data to train background model.For the rest of the execution it will be assumed that the pipeline is stateless.\n",
+ "There's no data to train background model.For the rest of the execution it will be assumed that the pipeline is stateless.\n",
+ "There's no data to train background model.For the rest of the execution it will be assumed that the pipeline is stateless.\n",
+ "There's no data to train background model.For the rest of the execution it will be assumed that the pipeline is stateless.\n",
+ "There's no data to train background model.For the rest of the execution it will be assumed that the pipeline is stateless.\n",
+ "There's no data to train background model.For the rest of the execution it will be assumed that the pipeline is stateless.\n",
+ "There's no data to train background model.For the rest of the execution it will be assumed that the pipeline is stateless.\n",
+ "There's no data to train background model.For the rest of the execution it will be assumed that the pipeline is stateless.\n",
+ "There's no data to train background model.For the rest of the execution it will be assumed that the pipeline is stateless.\n",
+ "There's no data to train background model.For the rest of the execution it will be assumed that the pipeline is stateless.\n"
+ ]
+ },
+ {
+ "data": {
+ "application/vnd.jupyter.widget-view+json": {
+ "model_id": "6556949a548c4f1da86073f0a9351109",
+ "version_major": 2,
+ "version_minor": 0
+ },
+ "text/plain": [
+ "Canvas(toolbar=Toolbar(toolitems=[('Home', 'Reset original view', 'home', 'home'), ('Back', 'Back to previous …"
+ ]
+ },
+ "metadata": {},
+ "output_type": "display_data"
+ }
+ ],
+ "source": [
+ "\"\"\"\n",
+ "vanilla_biometrics(\n",
+ " pipeline,\n",
+ " database,\n",
+ " dask_client,\n",
+ " groups,\n",
+ " output,\n",
+ " write_metadata_scores,\n",
+ " checkpoint,\n",
+ " dask_partition_size,\n",
+ " dask_n_workers,\n",
+ ")\n",
+ "\"\"\"\n",
+ "from sklearn.pipeline import make_pipeline\n",
+ "write_metadata_scores = False\n",
+ "checkpoint = False\n",
+ "dask_partition_size = None\n",
+ "dask_n_workers = 15\n",
+ "\n",
+ "### Preparing the pipeline\n",
+ "\n",
+ "fnmrs = []\n",
+ "for t in transformers:\n",
+ "\n",
+ " # Chain the Transformers together\n",
+ " transformer = make_pipeline(t, extractor_transformer)\n",
+ "\n",
+ " # Assemble the Vanilla Biometric pipeline and execute\n",
+ " pipeline = VanillaBiometricsPipeline(transformer, algorithm)\n",
+ "\n",
+ " execute_vanilla_biometrics(\n",
+ " pipeline,\n",
+ " database,\n",
+ " dask_client,\n",
+ " [\"dev\"],\n",
+ " output_path,\n",
+ " write_metadata_scores,\n",
+ " checkpoint,\n",
+ " dask_partition_size,\n",
+ " dask_n_workers,\n",
+ " allow_scoring_with_all_biometric_references=True\n",
+ " )\n",
+ " \n",
+ " scores_dev = os.path.join(output_path, \"scores-dev\")\n",
+ " \n",
+ " # Picking FNMR@FAR=0.001\n",
+ " neg, pos = bob.bio.base.score.load.split_four_column(scores_dev)\n",
+ " far_thres = bob.measure.far_threshold(neg, pos, 0.001)\n",
+ " fpr,fnr = bob.measure.fprfnr(neg, pos, far_thres)\n",
+ " fnmr_1 = round(1-fnr,2)\n",
+ " fnmrs.append(fnmr_1)\n",
+ " \n",
+ " \n",
+ "plot_faces(transformers, database, subplot_shape, fnmrs) \n"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": 5,
+ "metadata": {},
+ "outputs": [],
+ "source": [
+ "# Shutting down client\n",
+ "dask_client.shutdown()"
+ ]
+ }
+ ],
+ "metadata": {
+ "kernelspec": {
+ "display_name": "Python 3",
+ "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": "3.7.7"
+ }
+ },
+ "nbformat": 4,
+ "nbformat_minor": 4
+}
--
GitLab