Improve dataset card with paper link and task category

This PR improves the dataset card by:

- Adding a link to the associated paper: [ARKit LabelMaker: A New Scale for Indoor 3D Scene Understanding](https://huggingface.co/papers/2410.13924).
- Specifying the `task_categories` as `image-segmentation`.
- Clarifying the dataset's description to focus on ARKit LabelMaker itself.
- Adding relevant tags.

README.md

@@ -1,8 +1,16 @@
 ---
 license: creativeml-openrail-m
 viewer: false
+task_categories:
+- image-segmentation
+tags:
+- 3d
+- point-cloud
+- semantic-segmentation
+- indoor
+- arkit
 ---
-This is the dataset repo for training [Point Transformer](https://github.com/Pointcept/Pointcept) with the [ARKit LabelMaker](https://arxiv.org/abs/2410.13924) dataset. This dataset contains:
-1. the preprocessed data that is ready-to-use for Pointcept code base.
-2. the preprocessing code for official [ARKit LabelMaker dataset](https://huggingface.co/datasets/labelmaker/arkit_labelmaker).
-3. some experiment configurations.
+
+This dataset provides dense semantic annotations for the ARKitScenes dataset, creating a large-scale, real-world 3D dataset suitable for training 3D semantic segmentation models.  It was presented in the paper [ARKit LabelMaker: A New Scale for Indoor 3D Scene Understanding](https://huggingface.co/papers/2410.13924).
+
+This repository contains preprocessed data ready for use with the Pointcept codebase ([https://github.com/Pointcept/Pointcept](https://github.com/Pointcept/Pointcept)),  preprocessing code for the original ARKit LabelMaker dataset ([https://huggingface.co/datasets/labelmaker/arkit_labelmaker](https://huggingface.co/datasets/labelmaker/arkit_labelmaker)), and some experiment configurations.