Improve dataset card: Add task categories, paper link, relevant tags, and generation usage (#2)

- Improve dataset card: Add task categories, paper link, relevant tags, and generation usage (ba274943ea0cb23d9a0a9692e0d94044f77aec5e)


Co-authored-by: Niels Rogge <[email protected]>

README.md

@@ -1,11 +1,25 @@
 ---
 license: mit
+task_categories:
+- image-text-to-text
+language:
+- en
+tags:
+- vlm
+- spatial-reasoning
+- multi-view
+- vqa
+- cognition
 ---
-### Introduction
-Official dataset repo for paper: Reasoning Path and Latent State Analysis for Mulit-view Visual Spatial Reasoning: A Cognitive Science Perspective
-Github link: https://github.com/pittisl/ReMindView-Bench
 
-### Reconstructing the dataset
+# ReMindView-Bench Dataset
+
+[Paper](https://huggingface.co/papers/2512.02340) | [Code](https://github.com/pittisl/ReMindView-Bench)
+
+## Introduction
+ReMindView-Bench is a cognitively grounded benchmark for evaluating how Vision-Language Models (VLMs) construct, align, and maintain spatial mental models across complementary viewpoints. It addresses the struggle of current VLMs to maintain geometric coherence and cross-view consistency for spatial reasoning in multi-view settings by providing a fine-grained benchmark that isolates multi-view reasoning.
+
+## Reconstructing the dataset
 
 The dataset archive is split into 45GB parts to comply with the per-file limit. To rebuild the original tar after downloading all parts:
 
@@ -13,7 +27,34 @@ The dataset archive is split into 45GB parts to comply with the per-file limit.
 cat ReMindView-Bench.tar.part-* > ReMindView-Bench.tar
 ```
 
-### Dataset content
+## Dataset Generation
+
+To generate scenes, renders, and QA CSVs for the benchmark, follow these steps from the GitHub repository:
+
+1.  **Install Blender and Python Dependencies**:
+    You need to install Blender (headless is fine) and the Python dependencies used by Infinigen plus common packages. Run the scripts with Blender’s bundled Python or `blender --background --python <script> -- --flags` so `bpy` is available.
+
+2.  **Generate Scenes and Renders**:
+    From the repo root, generate scenes and renders:
+    ```bash
+    bash scene_generation.sh
+    ```
+    This script sweeps seeds 0–9 across five room types and writes scenes to `outputs/indoors/<ROOM>_<SEED>`, object-centric frames to `object_centric_view_frame_outputs/<ROOM>/<ROOM>_<SEED>`, and view-centric frames to `view_centric_view_frame_outputs/<ROOM>/<ROOM>_<SEED>`.
+
+3.  **Clean Empty/Invalid Views**:
+    ```bash
+    python clean_visual_data.py --dir_path object_centric_view_frame_outputs
+    ```
+    And the same for `view_centric_view_frame_outputs`.
+
+4.  **Produce QA CSVs**:
+    Choose one of `view_view`, `view_object`, `object_object`. For example:
+    ```bash
+    python ground_truth_generation.py --image_folder object_centric_view_frame_outputs --qa_type object_object
+    ```
+    The output CSV will be saved beside the image folder (e.g., `object_centric_view_frame_outputs/object_object_qa.csv`).
+
+## Dataset content
 
 VQA samples are stored in CSV files with the following columns: `folder_path` (scene/view folder), `query_type` (query relationship type), `query`, `ground_truth`, `choices`, `cross_frame` (whether cross frame reasoning is necessary), `perspective_changing` (whether requiring perspective changing), and `object_num` (object number in all frames).
 
@@ -27,7 +68,7 @@ Example row:
 - `perspective_changing`: `False`
 - `object_num`: `18` 
 
-### Sample scenes
+## Sample scenes
 
 Below are several example renders from ReMindView-Bench showing indoor layouts and object detail captured in the benchmark.
 Example 1:
@@ -84,4 +125,4 @@ Example 8:
 
 - Query: If you are positioned where the white small kitchen cabinet is, facing the same direction of the white small kitchen cabinet and then turn left, which object would be in the front of the dining table from this view direction?
 - Choice: A. wineglass, B. pot, C. chair
-- Answer: C. chair
+- Answer: C. chair