{ "cells": [ { "cell_type": "markdown", "id": "cab2d0f1", "metadata": {}, "source": [ "# **Trajectory** back on original video\n", "\n", "IDEA:\n", "- take the processed positions\n", "- apply the inverse of the homography\n", "- print on the video frame by frame all points before the current frame" ] }, { "cell_type": "code", "execution_count": 75, "id": "4cd7f188", "metadata": {}, "outputs": [], "source": [ "from pathlib import Path\n", "import cv2\n", "import csv\n", "\n", "import numpy as np\n", "import pandas as pd" ] }, { "cell_type": "code", "execution_count": 76, "id": "060e2535", "metadata": {}, "outputs": [], "source": [ "VIDEO_NUMBER = \"2\"\n", "PROJECT_ROOT = Path().resolve().parent.parent\n", "INPUT_VIDEO_PATH = str(\n", " PROJECT_ROOT\n", " / \"data\"\n", " / f\"recording_{VIDEO_NUMBER}\"\n", " / f\"Recording_{VIDEO_NUMBER}.mp4\"\n", ")\n", "TRASFORMED_CSV_PATH = str(\n", " PROJECT_ROOT\n", " / \"data\"\n", " / \"auxiliary_data\"\n", " / \"reconstructed_positions\"\n", " / f\"Transformed_positions_processed_TEST_{VIDEO_NUMBER}.csv\"\n", ")\n", "LANE_CSV_PATH = str(\n", " PROJECT_ROOT\n", " / \"data\"\n", " / \"auxiliary_data\"\n", " / \"lane_points\"\n", " / f\"Lane_points_{VIDEO_NUMBER}.csv\"\n", ")\n", "\n", "OUTPUT_VIDEO_PATH = str(\n", " PROJECT_ROOT\n", " / \"data\"\n", " / f\"recording_{VIDEO_NUMBER}\"\n", " / f\"Tracked_output_TEST_{VIDEO_NUMBER}.mp4\"\n", ")\n", "OUTPUT_CSV_PATH = str(\n", " PROJECT_ROOT\n", " / \"data\"\n", " / \"auxiliary_data\"\n", " / \"circle_positions\"\n", " / f\"Ball_lower_point_raw_TEST_{VIDEO_NUMBER}.csv\"\n", ")" ] }, { "cell_type": "code", "execution_count": 77, "id": "72b93c45", "metadata": {}, "outputs": [], "source": [ "lane_df = pd.read_csv(LANE_CSV_PATH)\n", "homographies = {}\n", "width_px, height_px = 106, 1829 # your real-world-to-pixel scale\n", "dst = np.array(\n", " [[0, height_px], [width_px, height_px], [width_px, 0], [0, 0]], dtype=np.float32\n", ")\n", "\n", "for _, r in lane_df.iterrows():\n", " frame_id = int(r[\"Frame\"])\n", " src = np.array(\n", " [\n", " [r[\"bottom_left_x\"], r[\"bottom_left_y\"]],\n", " [r[\"bottom_right_x\"], r[\"bottom_right_y\"]],\n", " [r[\"up_right_x\"], r[\"up_right_y\"]],\n", " [r[\"up_left_x\"], r[\"up_left_y\"]],\n", " ],\n", " dtype=np.float32,\n", " )\n", " H, _ = cv2.findHomography(dst, src)\n", " if H is not None:\n", " homographies[frame_id] = H" ] }, { "cell_type": "code", "execution_count": 78, "id": "47f2a59d", "metadata": {}, "outputs": [], "source": [ "positions = {}\n", "with open(TRASFORMED_CSV_PATH, newline=\"\") as f:\n", " reader = csv.reader(f)\n", " next(reader) # skip header\n", " for frame_str, x_str, y_str in reader:\n", " fid = int(frame_str)\n", " if x_str and y_str:\n", " positions[fid] = (float(x_str), float(y_str))\n", " else:\n", " positions[fid] = None" ] }, { "cell_type": "code", "execution_count": 79, "id": "dab28bc3", "metadata": {}, "outputs": [], "source": [ "cap = cv2.VideoCapture(str(INPUT_VIDEO_PATH))\n", "if not cap.isOpened():\n", " raise RuntimeError(f\"Cannot open {INPUT_VIDEO_PATH!r}\")\n", "\n", "w = int(cap.get(cv2.CAP_PROP_FRAME_WIDTH))\n", "h = int(cap.get(cv2.CAP_PROP_FRAME_HEIGHT))\n", "fps = cap.get(cv2.CAP_PROP_FPS)\n", "\n", "fourcc = cv2.VideoWriter_fourcc(*\"mp4v\")\n", "out = cv2.VideoWriter(str(OUTPUT_VIDEO_PATH), fourcc, fps, (w, h))" ] }, { "cell_type": "code", "execution_count": 80, "id": "27d15335", "metadata": {}, "outputs": [], "source": [ "trajectory = [] # list of (x,y) in real-world space\n", "last_H = None\n", "frame_idx = 0\n", "saved_pts = []\n", "\n", "while True:\n", " ret, frame = cap.read()\n", " if not ret:\n", " break\n", "\n", " # Update homography if available\n", " if frame_idx in homographies:\n", " last_H = homographies[frame_idx]\n", "\n", " # Append new real-world point\n", " pt = positions.get(frame_idx)\n", " if pt is not None:\n", " trajectory.append(pt)\n", "\n", " # If we have a homography and trajectory, project & draw\n", " if last_H is not None and trajectory:\n", " # Prepare points in Nx1x2 shape\n", " pts_world = np.array(trajectory, dtype=np.float32).reshape(-1, 1, 2)\n", " pts_img = cv2.perspectiveTransform(pts_world, last_H).reshape(-1, 2)\n", " pts_int = np.round(pts_img).astype(np.int32)\n", "\n", " # Draw full trajectory polyline\n", " cv2.polylines(frame, [pts_int], isClosed=False, color=(0, 0, 255), thickness=2)\n", "\n", " # Draw most recent point as a filled circle\n", " x0, y0 = tuple(pts_int[-1])\n", " cv2.circle(frame, (x0, y0), radius=4, color=(0, 0, 255), thickness=-1)\n", " saved_pts.append([frame_idx, x0, y0])\n", "\n", " out.write(frame)\n", " frame_idx += 1" ] }, { "cell_type": "code", "execution_count": 81, "id": "e63549c3", "metadata": {}, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "Tracking video saved to: /home/davic/projects/IACV_project/bowling-analysis/data/recording_2/Tracked_output_TEST_2.mp4\n" ] } ], "source": [ "cap.release()\n", "out.release()\n", "print(f\"Tracking video saved to: {OUTPUT_VIDEO_PATH}\")\n", "\n", "# Save the raw projected points\n", "df_out = pd.DataFrame(saved_pts, columns=[\"frame\", \"x\", \"y\"])\n", "# df_out.to_csv(OUTPUT_CSV_PATH, index=False)\n", "# print(f\"Projected points saved to: {OUTPUT_CSV_PATH}\")" ] } ], "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": "3.12.3" } }, "nbformat": 4, "nbformat_minor": 5 }