Convert to Parquet format (remove loading script)

README.md

@@ -37,20 +37,21 @@ The dataset currently includes:
 - **3 questions per scene**: one for each image
 - **Answer matrix**: the answer of every image to every question
 
-Data is stored in `dataset_1k_720p_2/image_mapping_with_questions.csv` with images in `dataset_1k_720p_2/images/`.
+Data is stored in **Parquet format** (`data/train.parquet`) with images in `dataset_1k_720p_2/images/`.
 
 ## Hugging Face Dataset Viewer
 
-This repository is configured as a Hugging Face dataset. The **Preview** tab will automatically:
+This repository uses the standard **Parquet format** (no custom loading script required). The **Preview** tab will automatically:
 
-1. Use `MMB_Dataset.py` to load the CSV from `dataset_1k_720p_2/`
-2. Resolve image paths to display thumbnails
-3. Show all fields including:
+1. Load the Parquet file from `data/train.parquet`
+2. Display all fields including:
    - Scene IDs
-   - Three images per example (original + 2 counterfactuals)
+   - Image filenames (original + 2 counterfactuals)
    - Questions and difficulties
    - Complete answer matrix (9 fields)
 
+**Note:** Image columns contain filenames (e.g., `scene_0000_original.png`). The actual images are stored in `dataset_1k_720p_2/images/`.
+
 ### Loading from Python
 
 After pushing this repository to the Hub, load it with:
@@ -58,21 +59,24 @@ After pushing this repository to the Hub, load it with:
 ```python
 from datasets import load_dataset
 
-ds = load_dataset("scholo/MMB_dataset", split="train", trust_remote_code=True)
+ds = load_dataset("scholo/MMB_dataset", split="train")
 print(ds[0])
 ```
 
+No `trust_remote_code=True` needed since we use standard Parquet format!
+
 ## Directory Structure
 
 ```
 MMB-Dataset/
-├── MMB_dataset.py              # Hugging Face dataset loading script
 ├── README.md                   # This file
 ├── .gitattributes              # Git LFS configuration for images
+├── data/                       # Dataset files (Parquet format)
+│   └── train.parquet           # Main dataset file
 ├── dataset_1k_720p_2/          # Current dataset run
-│   ├── images/                 # All PNG images
+│   ├── images/                 # All PNG images (referenced by Parquet)
 │   ├── scenes/                 # JSON scene descriptions (reference)
-│   ├── image_mapping_with_questions.csv  # Main data file
+│   ├── image_mapping_with_questions.csv  # Original CSV (source)
 │   ├── checkpoint.json         # Run metadata
 │   └── run_metadata.json       # Run metadata
 ```

data/train.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1a39ca89d19e06768c08b1b217dd5814336a570c76898aef79bca4c44a72213b
+size 49238