nb to surgically fix the timestamps in the parquet file

_dev/fix_timestamp.ipynb

@@ -0,0 +1,345 @@
+{
+ "cells": [
+  {
+   "cell_type": "code",
+   "execution_count": 14,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "import pandas as pd\n",
+    "import cv2\n",
+    "from tqdm import tqdm"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 15,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "dataframes = [\n",
+    "    \"../data/dataset-unfiltered.parquet\",\n",
+    "    \"../data/dataset.parquet\",\n",
+    "]\n",
+    "df_path = dataframes[1]"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 16,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "df = pd.read_parquet(df_path)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 17,
+   "metadata": {},
+   "outputs": [
+    {
+     "data": {
+      "text/html": [
+       "<div>\n",
+       "<style scoped>\n",
+       "    .dataframe tbody tr th:only-of-type {\n",
+       "        vertical-align: middle;\n",
+       "    }\n",
+       "\n",
+       "    .dataframe tbody tr th {\n",
+       "        vertical-align: top;\n",
+       "    }\n",
+       "\n",
+       "    .dataframe thead th {\n",
+       "        text-align: right;\n",
+       "    }\n",
+       "</style>\n",
+       "<table border=\"1\" class=\"dataframe\">\n",
+       "  <thead>\n",
+       "    <tr style=\"text-align: right;\">\n",
+       "      <th></th>\n",
+       "      <th>video_id</th>\n",
+       "      <th>frame_idx</th>\n",
+       "      <th>timestamp</th>\n",
+       "      <th>base64_image</th>\n",
+       "      <th>dim_0</th>\n",
+       "      <th>dim_1</th>\n",
+       "      <th>dim_2</th>\n",
+       "      <th>dim_3</th>\n",
+       "      <th>dim_4</th>\n",
+       "      <th>dim_5</th>\n",
+       "      <th>...</th>\n",
+       "      <th>dim_502</th>\n",
+       "      <th>dim_503</th>\n",
+       "      <th>dim_504</th>\n",
+       "      <th>dim_505</th>\n",
+       "      <th>dim_506</th>\n",
+       "      <th>dim_507</th>\n",
+       "      <th>dim_508</th>\n",
+       "      <th>dim_509</th>\n",
+       "      <th>dim_510</th>\n",
+       "      <th>dim_511</th>\n",
+       "    </tr>\n",
+       "  </thead>\n",
+       "  <tbody>\n",
+       "    <tr>\n",
+       "      <th>0</th>\n",
+       "      <td>8Ilh1ewceco</td>\n",
+       "      <td>145</td>\n",
+       "      <td>5.0</td>\n",
+       "      <td>b'/9j/4AAQSkZJRgABAQAAAQABAAD/2wBDAAgGBgcGBQgH...</td>\n",
+       "      <td>0.009040</td>\n",
+       "      <td>0.003338</td>\n",
+       "      <td>0.029684</td>\n",
+       "      <td>-0.033058</td>\n",
+       "      <td>0.040864</td>\n",
+       "      <td>-0.006447</td>\n",
+       "      <td>...</td>\n",
+       "      <td>0.033575</td>\n",
+       "      <td>-0.019076</td>\n",
+       "      <td>0.047166</td>\n",
+       "      <td>-0.010574</td>\n",
+       "      <td>-0.018608</td>\n",
+       "      <td>-0.013465</td>\n",
+       "      <td>-0.020017</td>\n",
+       "      <td>0.086240</td>\n",
+       "      <td>-0.029653</td>\n",
+       "      <td>0.035949</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>1</th>\n",
+       "      <td>8Ilh1ewceco</td>\n",
+       "      <td>290</td>\n",
+       "      <td>10.0</td>\n",
+       "      <td>b'/9j/4AAQSkZJRgABAQAAAQABAAD/2wBDAAgGBgcGBQgH...</td>\n",
+       "      <td>0.004891</td>\n",
+       "      <td>0.006527</td>\n",
+       "      <td>0.004417</td>\n",
+       "      <td>-0.000323</td>\n",
+       "      <td>0.006400</td>\n",
+       "      <td>-0.024191</td>\n",
+       "      <td>...</td>\n",
+       "      <td>-0.043122</td>\n",
+       "      <td>-0.010695</td>\n",
+       "      <td>0.005672</td>\n",
+       "      <td>0.000172</td>\n",
+       "      <td>-0.014442</td>\n",
+       "      <td>-0.014647</td>\n",
+       "      <td>-0.016840</td>\n",
+       "      <td>0.100285</td>\n",
+       "      <td>0.013794</td>\n",
+       "      <td>0.015046</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>2</th>\n",
+       "      <td>8Ilh1ewceco</td>\n",
+       "      <td>435</td>\n",
+       "      <td>15.0</td>\n",
+       "      <td>b'/9j/4AAQSkZJRgABAQAAAQABAAD/2wBDAAgGBgcGBQgH...</td>\n",
+       "      <td>-0.022159</td>\n",
+       "      <td>0.020703</td>\n",
+       "      <td>-0.021607</td>\n",
+       "      <td>-0.019721</td>\n",
+       "      <td>-0.006067</td>\n",
+       "      <td>-0.035070</td>\n",
+       "      <td>...</td>\n",
+       "      <td>-0.017047</td>\n",
+       "      <td>-0.018341</td>\n",
+       "      <td>-0.006733</td>\n",
+       "      <td>-0.007040</td>\n",
+       "      <td>-0.008368</td>\n",
+       "      <td>0.009755</td>\n",
+       "      <td>-0.045662</td>\n",
+       "      <td>0.116601</td>\n",
+       "      <td>-0.000572</td>\n",
+       "      <td>-0.000985</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>3</th>\n",
+       "      <td>8Ilh1ewceco</td>\n",
+       "      <td>580</td>\n",
+       "      <td>20.0</td>\n",
+       "      <td>b'/9j/4AAQSkZJRgABAQAAAQABAAD/2wBDAAgGBgcGBQgH...</td>\n",
+       "      <td>-0.015903</td>\n",
+       "      <td>0.033545</td>\n",
+       "      <td>0.009257</td>\n",
+       "      <td>-0.033540</td>\n",
+       "      <td>0.010586</td>\n",
+       "      <td>-0.028067</td>\n",
+       "      <td>...</td>\n",
+       "      <td>-0.016532</td>\n",
+       "      <td>0.012388</td>\n",
+       "      <td>0.020868</td>\n",
+       "      <td>-0.012635</td>\n",
+       "      <td>0.010914</td>\n",
+       "      <td>0.009203</td>\n",
+       "      <td>-0.010078</td>\n",
+       "      <td>0.063971</td>\n",
+       "      <td>-0.038024</td>\n",
+       "      <td>0.025840</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>4</th>\n",
+       "      <td>8Ilh1ewceco</td>\n",
+       "      <td>725</td>\n",
+       "      <td>25.0</td>\n",
+       "      <td>b'/9j/4AAQSkZJRgABAQAAAQABAAD/2wBDAAgGBgcGBQgH...</td>\n",
+       "      <td>-0.010193</td>\n",
+       "      <td>-0.012323</td>\n",
+       "      <td>0.023012</td>\n",
+       "      <td>-0.015893</td>\n",
+       "      <td>0.047041</td>\n",
+       "      <td>0.050783</td>\n",
+       "      <td>...</td>\n",
+       "      <td>-0.024013</td>\n",
+       "      <td>-0.009684</td>\n",
+       "      <td>-0.035481</td>\n",
+       "      <td>-0.024743</td>\n",
+       "      <td>-0.009812</td>\n",
+       "      <td>0.035214</td>\n",
+       "      <td>-0.008957</td>\n",
+       "      <td>0.124215</td>\n",
+       "      <td>-0.012410</td>\n",
+       "      <td>0.040907</td>\n",
+       "    </tr>\n",
+       "  </tbody>\n",
+       "</table>\n",
+       "<p>5 rows × 516 columns</p>\n",
+       "</div>"
+      ],
+      "text/plain": [
+       "      video_id  frame_idx  timestamp   \n",
+       "0  8Ilh1ewceco        145        5.0  \\\n",
+       "1  8Ilh1ewceco        290       10.0   \n",
+       "2  8Ilh1ewceco        435       15.0   \n",
+       "3  8Ilh1ewceco        580       20.0   \n",
+       "4  8Ilh1ewceco        725       25.0   \n",
+       "\n",
+       "                                        base64_image     dim_0     dim_1   \n",
+       "0  b'/9j/4AAQSkZJRgABAQAAAQABAAD/2wBDAAgGBgcGBQgH...  0.009040  0.003338  \\\n",
+       "1  b'/9j/4AAQSkZJRgABAQAAAQABAAD/2wBDAAgGBgcGBQgH...  0.004891  0.006527   \n",
+       "2  b'/9j/4AAQSkZJRgABAQAAAQABAAD/2wBDAAgGBgcGBQgH... -0.022159  0.020703   \n",
+       "3  b'/9j/4AAQSkZJRgABAQAAAQABAAD/2wBDAAgGBgcGBQgH... -0.015903  0.033545   \n",
+       "4  b'/9j/4AAQSkZJRgABAQAAAQABAAD/2wBDAAgGBgcGBQgH... -0.010193 -0.012323   \n",
+       "\n",
+       "      dim_2     dim_3     dim_4     dim_5  ...   dim_502   dim_503   dim_504   \n",
+       "0  0.029684 -0.033058  0.040864 -0.006447  ...  0.033575 -0.019076  0.047166  \\\n",
+       "1  0.004417 -0.000323  0.006400 -0.024191  ... -0.043122 -0.010695  0.005672   \n",
+       "2 -0.021607 -0.019721 -0.006067 -0.035070  ... -0.017047 -0.018341 -0.006733   \n",
+       "3  0.009257 -0.033540  0.010586 -0.028067  ... -0.016532  0.012388  0.020868   \n",
+       "4  0.023012 -0.015893  0.047041  0.050783  ... -0.024013 -0.009684 -0.035481   \n",
+       "\n",
+       "    dim_505   dim_506   dim_507   dim_508   dim_509   dim_510   dim_511  \n",
+       "0 -0.010574 -0.018608 -0.013465 -0.020017  0.086240 -0.029653  0.035949  \n",
+       "1  0.000172 -0.014442 -0.014647 -0.016840  0.100285  0.013794  0.015046  \n",
+       "2 -0.007040 -0.008368  0.009755 -0.045662  0.116601 -0.000572 -0.000985  \n",
+       "3 -0.012635  0.010914  0.009203 -0.010078  0.063971 -0.038024  0.025840  \n",
+       "4 -0.024743 -0.009812  0.035214 -0.008957  0.124215 -0.012410  0.040907  \n",
+       "\n",
+       "[5 rows x 516 columns]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    }
+   ],
+   "source": [
+    "display(df.head())"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 18,
+   "metadata": {},
+   "outputs": [
+    {
+     "data": {
+      "text/plain": [
+       "60844"
+      ]
+     },
+     "execution_count": 18,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "# wierdly all the timestamps are rounded to the nearest second o.O\n",
+    "(df[\"timestamp\"].apply(lambda x: x % 1) == 0).sum()"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 19,
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "100%|██████████| 1417/1417 [00:02<00:00, 626.05it/s]\n"
+     ]
+    }
+   ],
+   "source": [
+    "video_fps = {}\n",
+    "for video_id in tqdm(df.video_id.unique()):\n",
+    "    video_path = f\"../data/videos/{video_id}.mp4\"\n",
+    "    cap = cv2.VideoCapture(str(video_path))\n",
+    "    num_video_frames = int(cap.get(cv2.CAP_PROP_FRAME_COUNT))\n",
+    "    fps = cap.get(cv2.CAP_PROP_FPS)\n",
+    "    video_fps[video_id] = fps"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 20,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "def correct_timestamp(row):\n",
+    "    # pd.Series([1, 2], index=['foo', 'bar'])\n",
+    "    video_id = row[\"video_id\"]\n",
+    "    frame_idx = row[\"frame_idx\"]\n",
+    "    fps = video_fps[video_id]\n",
+    "    timestamp = frame_idx / fps\n",
+    "    return timestamp\n",
+    "\n",
+    "\n",
+    "df[\"timestamp\"] = df.apply(correct_timestamp, axis=1)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 22,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "df.to_parquet(df_path, index=False)"
+   ]
+  }
+ ],
+ "metadata": {
+  "kernelspec": {
+   "display_name": "visual-content-search-over-videos",
+   "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": "3.9.16"
+  },
+  "orig_nbformat": 4
+ },
+ "nbformat": 4,
+ "nbformat_minor": 2
+}