Update README.md

README.md

@@ -50,42 +50,4 @@ Example use cases:
 
 Do **not** use this dataset as training data. If you require a trainable dataset, you must substitute animations that are licensed for ML use.
 
-Example usage:
-```python
-# After downloading & unzipping:
-# Synthetic_reasoning_dataset/
-#   anomaly_videos/...
-#   follow_videos/...
-#   spatial_colored_videos/...
-#   spatial_videos/...
-
-from datasets.synthetic_reasoning_dataset import (
-    SyntheticReasoningDataset,
-    build_index_dataframe,
-)
-
-root = "path/to/Synthetic_reasoning_dataset"
-
-# 1) Quick index as a DataFrame (great for sanity checks)
-df = build_index_dataframe(root)
-print(df.head())
-
-# 2) PyTorch-style dataset without decoding (fast)
-ds = SyntheticReasoningDataset(root, tasks=None, decode=False)
-path, label_id, meta = ds[0]
-print(path, label_id, meta)
-
-# 3) With decoding (requires torchvision); sample every 2nd frame, cap at 64 frames
-ds_decoded = SyntheticReasoningDataset(root, tasks=["anomaly", "follow"], decode=True, sample_stride=2, max_frames=64)
-video, label_id, meta = ds_decoded[0]
-print(video.shape, label_id, meta)
-
-# 4) Label spaces
-from datasets.synthetic_reasoning_dataset import LABELS_PER_TASK, LABEL2ID
-print(LABELS_PER_TASK)
-print(LABEL2ID)
-
-```
-
----
-
+For testing with vision-LLMs see our [GitHub repo](https://github.com/pascalbenschopTU/VLLM_AnomalyRecognition).