Fix description in object dection notebook

Author: cgeller

section_2_sensor_data_processing/5_object_detection.ipynb

@@ -255,7 +255,7 @@
    "metadata": {},
    "source": [
     "### Task 2: Training / Validation / Test Split\n",
-    "Training neural networks requires several preprocessing steps. In addition to steps as data augmentation or normalization, data has to be loaded and distributed to training, validation, and test datasets. For the split into these three datasets one often finds a 70/20/10 split in literature, so you should apply the same split to the 200 samples from the KITTI dataset.\n",
+    "Training neural networks requires several preprocessing steps. In addition to steps as data augmentation or normalization, data has to be loaded and distributed to training, validation, and test datasets. For the split into these three datasets one often finds a 70/20/10 split in literature, so you should apply the same split to the 100 samples from the KITTI dataset.\n",
     "\n",
     "Your task is to create the three datasets with the class `SimpleDataGenerator` and its constructor:\n",
     "```\n",
@@ -338,7 +338,7 @@
    "cell_type": "markdown",
    "metadata": {},
    "source": [
-    "Note: The batch size is set to 2, because of large memory consumption of LiDAR point clouds. Hence, only small batch sizes are typically feasible."
+    "Note: The batch size is set to 4, because of large memory consumption of LiDAR point clouds. Hence, only small batch sizes are typically feasible."
    ]
   },
   {