From 981b0f7e552bfd85c48ee3362d597da43c5c6c50 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?OTH=C3=89GUY=20Marion?= <marion.otheguy@imt-atlantique.fr>
Date: Fri, 14 Mar 2025 09:32:32 +0000
Subject: [PATCH] Upload New File
---
preprocessing.ipynb | 279 ++++++++++++++++++++++++++++++++++++++++++++
1 file changed, 279 insertions(+)
create mode 100644 preprocessing.ipynb
diff --git a/preprocessing.ipynb b/preprocessing.ipynb
new file mode 100644
index 0000000..7de2d44
--- /dev/null
+++ b/preprocessing.ipynb
@@ -0,0 +1,279 @@
+{
+ "cells": [
+ {
+ "cell_type": "code",
+ "execution_count": null,
+ "metadata": {},
+ "outputs": [],
+ "source": [
+ "#Libs\n",
+ "import cv2\n",
+ "import numpy as np\n",
+ "import pandas as pd\n",
+ "from tabulate import tabulate\n",
+ "import matplotlib.pyplot as plt"
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "metadata": {},
+ "source": [
+ "# Methods"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": null,
+ "metadata": {},
+ "outputs": [],
+ "source": [
+ "def split_on_nan_row(recording):\n",
+ " # Split the DataFrame into subdatasets\n",
+ " subdatasets = []\n",
+ " current_subdataset = []\n",
+ "\n",
+ " for index, row in recording.iterrows():\n",
+ " if pd.isna(row['x1']) and pd.isna(row['y1']) and pd.isna(row['x2']) and pd.isna(row['y2']):\n",
+ " if current_subdataset:\n",
+ " subdatasets.append(pd.DataFrame(current_subdataset))\n",
+ " current_subdataset = []\n",
+ " elif pd.isna(row['x1']) and pd.isna(row['y1']) and row['x2'] and row['y2']:\n",
+ " if current_subdataset:\n",
+ " subdatasets.append(pd.DataFrame(current_subdataset))\n",
+ " current_subdataset = []\n",
+ " elif row['x1'] and row['y1'] and pd.isna(row['x2']) and pd.isna(row['y2']):\n",
+ " if current_subdataset:\n",
+ " subdatasets.append(pd.DataFrame(current_subdataset))\n",
+ " current_subdataset = []\n",
+ " else:\n",
+ " if index > 0:\n",
+ " prev_row = recording.iloc[index - 1]\n",
+ " if not (pd.isna(prev_row['x1']) and pd.isna(prev_row['y1']) and pd.isna(prev_row['x2']) and pd.isna(prev_row['y2'])):\n",
+ " current_subdataset.append(row)\n",
+ "\n",
+ " # Append the last subdataset if it exists\n",
+ " if current_subdataset:\n",
+ " subdatasets.append(pd.DataFrame(current_subdataset))\n",
+ "\n",
+ " for i in range(len(subdatasets)):\n",
+ " if len(subdatasets[i]) < 10:\n",
+ " subdatasets.pop(i)\n",
+ " break\n",
+ " return subdatasets"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": null,
+ "metadata": {},
+ "outputs": [],
+ "source": [
+ "def get_average_points(subdatasets):\n",
+ " average_points = []\n",
+ "\n",
+ " for i in subdatasets:\n",
+ " x1 = i['x1'].mean() \n",
+ " y1 = i['y1'].mean()\n",
+ " x2 = i['x2'].mean()\n",
+ " y2 = i['y2'].mean()\n",
+ " average_points.append((x1, y1, x2, y2))\n",
+ "\n",
+ " return average_points"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": null,
+ "metadata": {},
+ "outputs": [],
+ "source": [
+ "def correct_points_orientation(list_of_points, label):\n",
+ " average_points_corrected = []\n",
+ " for i in range(len(list_of_points)):\n",
+ " x1, y1, x2, y2 = list_of_points[i]\n",
+ " if x1 < x2:\n",
+ " x1, x2 = x2, x1\n",
+ " y1, y2 = y2, y1\n",
+ " average_points_corrected.append((x1, y1, x2, y2, label[i]))\n",
+ " return average_points_corrected"
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "metadata": {},
+ "source": [
+ "# Variables"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": null,
+ "metadata": {},
+ "outputs": [],
+ "source": [
+ "# Angular error\n",
+ "angle_error = np.radians(0.25) # 0.25°\n",
+ "\n",
+ "# Define the positions of the eyes (world coordinates)\n",
+ "coord_left_eye = np.array([-3.25, 0, 0])\n",
+ "coord_right_eye = np.array([3.25, 0, 0])"
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "metadata": {},
+ "source": [
+ "# Recordings"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": null,
+ "metadata": {},
+ "outputs": [],
+ "source": [
+ "label1 = ['infini', '25c', '25up', '25r', '50c', '50up', '50r', '75c', '75up', '75r', '200c', '200up', '200r' ]\n",
+ "\n",
+ "recording1_left = pd.read_csv('./recording1/centroids_left.csv')\n",
+ "recording1_right = pd.read_csv('./recording1/centroids_right.csv')\n",
+ "\n",
+ "# Retenir seulement les images non noirs\n",
+ "subdatasets1_left = split_on_nan_row(recording1_left)\n",
+ "subdatasets1_right = split_on_nan_row(recording1_right)\n",
+ "\n",
+ "# Recupérer les points moyens pour les douze points de fixations\n",
+ "average_points_left1 = get_average_points(subdatasets1_left)\n",
+ "average_points_right1 = get_average_points(subdatasets1_right)\n",
+ "\n",
+ "# Verifier qu'il n'y a pas d'échanges entre les centroides 1 et 2\n",
+ "average_points_left_corrected1, average_points_right_corrected1 = correct_points_orientation(average_points_left1, label1), correct_points_orientation(average_points_right1, label1)\n",
+ "\n",
+ "xy_eye_left1 = np.column_stack(([(average_points_left_corrected1[i][0] + average_points_left_corrected1[i][2])/2 for i in range(len(average_points_left_corrected1))], [(average_points_left_corrected1[i][1] + average_points_left_corrected1[i][3])/2 for i in range(len(average_points_left_corrected1))]))\n",
+ "xy_eye_right1 = np.column_stack(([(average_points_right_corrected1[i][0] + average_points_right_corrected1[i][2])/2 for i in range(len(average_points_right_corrected1))], [(average_points_right_corrected1[i][1] + average_points_right_corrected1[i][3])/2 for i in range(len(average_points_right_corrected1))]))"
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "metadata": {},
+ "source": [
+ "## Recording 2"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": null,
+ "metadata": {},
+ "outputs": [],
+ "source": [
+ "label2 = ['infini', '25c', '50c', '75c', '200c', '25up', '50up', '75up', '200up', '25r', '50r', '75r', '200r' ]\n",
+ "\n",
+ "recording2_left = pd.read_csv('./recording2/centroids_left.csv')\n",
+ "recording2_left = recording2_left.drop([11474])\n",
+ "recording2_left = recording2_left.reset_index(drop=True)\n",
+ "\n",
+ "recording2_right = pd.read_csv('./.recording2/centroids_right.csv')\n",
+ "recording2_right = recording2_right.drop([319, 5926, 7611, 10201, 11500, 12791])\n",
+ "recording2_right = recording2_right.reset_index(drop=True)\n",
+ "\n",
+ "subdatasets2_left = split_on_nan_row(recording2_left)\n",
+ "subdatasets2_right = split_on_nan_row(recording2_right)\n",
+ "\n",
+ "average_points_left2 = get_average_points(subdatasets2_left)\n",
+ "average_points_right2 = get_average_points(subdatasets2_right)\n",
+ "\n",
+ "average_points_left_corrected2, average_points_right_corrected2 = correct_points_orientation(average_points_left2, label2), correct_points_orientation(average_points_right2, label2)\n",
+ "\n",
+ "xy_eye_left2 = np.column_stack(([(average_points_left_corrected2[i][0] + average_points_left_corrected2[i][2])/2 for i in range(len(average_points_left_corrected2))], [(average_points_left_corrected2[i][1] + average_points_left_corrected2[i][3])/2 for i in range(len(average_points_left_corrected2))]))\n",
+ "xy_eye_right2 = np.column_stack(([(average_points_right_corrected2[i][0] + average_points_right_corrected2[i][2])/2 for i in range(len(average_points_right_corrected2))], [(average_points_right_corrected2[i][1] + average_points_right_corrected2[i][3])/2 for i in range(len(average_points_right_corrected2))]))"
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "metadata": {},
+ "source": [
+ "## Recording 3"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": null,
+ "metadata": {},
+ "outputs": [],
+ "source": [
+ "label3 = ['infini', '75c', '75up', '50up', '75r', '50r', '200c', '25c', '25r', '50c', '200r', '25up', '200up']\n",
+ "\n",
+ "recording3_left = pd.read_csv('./recording3/centroids_left.csv')\n",
+ "recording3_right = pd.read_csv('./recording3/centroids_right.csv')\n",
+ "recording3_right = recording3_right.drop([9268, 10181, 11001, 12687])\n",
+ "recording3_right = recording3_right.reset_index(drop=True)\n",
+ "\n",
+ "subdatasets3_left = split_on_nan_row(recording3_left)\n",
+ "subdatasets3_right = split_on_nan_row(recording3_right)\n",
+ "\n",
+ "average_points_left3 = get_average_points(subdatasets3_left)\n",
+ "average_points_right3 = get_average_points(subdatasets3_right)\n",
+ "\n",
+ "average_points_left_corrected3, average_points_right_corrected3 = correct_points_orientation(average_points_left3, label3), correct_points_orientation(average_points_right3, label3)\n",
+ "\n",
+ "xy_eye_left3 = np.column_stack(([(average_points_left_corrected3[i][0] + average_points_left_corrected3[i][2])/2 for i in range(len(average_points_left_corrected3))], [(average_points_left_corrected3[i][1] + average_points_left_corrected3[i][3])/2 for i in range(len(average_points_left_corrected3))]))\n",
+ "xy_eye_right3 = np.column_stack(([(average_points_right_corrected3[i][0] + average_points_right_corrected3[i][2])/2 for i in range(len(average_points_right_corrected3))], [(average_points_right_corrected3[i][1] + average_points_right_corrected3[i][3])/2 for i in range(len(average_points_right_corrected3))]))"
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "metadata": {},
+ "source": [
+ "## Recording 4"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": null,
+ "metadata": {},
+ "outputs": [],
+ "source": [
+ "label4 = ['infini', '25up', '200c', '200up', '75c', '25c', '50c', '200r', '25r', '75r', '75up', '50up', '50r' ]\n",
+ "\n",
+ "recording4_left = pd.read_csv('./recording4/centroids_left.csv')\n",
+ "recording4_left = recording4_left.drop([1971,1972, 4402, 6891,7982,9011, 10229, 10434, 12390, 13182])\n",
+ "recording4_left = recording4_left.reset_index(drop=True)\n",
+ "\n",
+ "recording4_right = pd.read_csv('./recording4/centroids_right.csv')\n",
+ "recording4_right = recording4_right.drop([1960, 4394, 8170, 10423, 13177])\n",
+ "recording4_right = recording4_right.reset_index(drop=True)\n",
+ "\n",
+ "subdatasets4_left = split_on_nan_row(recording4_left)\n",
+ "subdatasets4_right = split_on_nan_row(recording4_right)\n",
+ "\n",
+ "average_points_left4 = get_average_points(subdatasets4_left)\n",
+ "average_points_right4 = get_average_points(subdatasets4_right)\n",
+ "\n",
+ "average_points_left_corrected4, average_points_right_corrected4 = correct_points_orientation(average_points_left4, label4), correct_points_orientation(average_points_right4, label4)\n",
+ "\n",
+ "xy_eye_left4 = np.column_stack(([(average_points_left_corrected4[i][0] + average_points_left_corrected4[i][2])/2 for i in range(len(average_points_left_corrected4))], [(average_points_left_corrected4[i][1] + average_points_left_corrected4[i][3])/2 for i in range(len(average_points_left_corrected4))]))\n",
+ "xy_eye_right4 = np.column_stack(([(average_points_right_corrected4[i][0] + average_points_right_corrected4[i][2])/2 for i in range(len(average_points_right_corrected4))], [(average_points_right_corrected4[i][1] + average_points_right_corrected4[i][3])/2 for i in range(len(average_points_right_corrected4))]))"
+ ]
+ }
+ ],
+ "metadata": {
+ "kernelspec": {
+ "display_name": "img_pro",
+ "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": "3.12.0"
+ }
+ },
+ "nbformat": 4,
+ "nbformat_minor": 2
+}
--
GitLab