Datasets:
Improve dataset card with description, links, and categories
#2
by
nielsr
HF Staff
- opened
README.md
CHANGED
|
@@ -1,4 +1,12 @@
|
|
| 1 |
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 2 |
dataset_info:
|
| 3 |
features:
|
| 4 |
- name: relative_path
|
|
@@ -21,3 +29,30 @@ configs:
|
|
| 21 |
- split: train
|
| 22 |
path: data/train-*
|
| 23 |
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
---
|
| 2 |
+
license: mit
|
| 3 |
+
task_categories:
|
| 4 |
+
- video-classification
|
| 5 |
+
tags:
|
| 6 |
+
- temporal-reasoning
|
| 7 |
+
- video-understanding
|
| 8 |
+
- benchmark
|
| 9 |
+
- vision-language
|
| 10 |
dataset_info:
|
| 11 |
features:
|
| 12 |
- name: relative_path
|
|
|
|
| 29 |
- split: train
|
| 30 |
path: data/train-*
|
| 31 |
---
|
| 32 |
+
|
| 33 |
+
# SpookyBench: A Benchmark for Purely Temporal Video Understanding
|
| 34 |
+
|
| 35 |
+
SpookyBench is a novel benchmark dataset designed to evaluate the ability of video-language models (VLMs) to understand purely temporal patterns, independent of spatial cues. The dataset consists of 451 videos across four categories: Text, Object Images, Dynamic Scenes, and Shapes. Each video appears as random noise in individual frames, but reveals meaningful content (words, objects, etc.) when viewed as a temporal sequence. This design exposes a critical limitation in current VLMs, which often heavily rely on spatial information and struggle to extract meaning from purely temporal sequences.
|
| 36 |
+
|
| 37 |
+
[Paper: Time Blindness: Why Video-Language Models Can't See What Humans Can?](https://huggingface.co/papers/2505.24867)
|
| 38 |
+
|
| 39 |
+
[Project Website: https://timeblindness.github.io/](https://timeblindness.github.io/)
|
| 40 |
+
|
| 41 |
+
The dataset contains 451 videos distributed as follows:
|
| 42 |
+
|
| 43 |
+
| **Category** | **Total Videos** | **Description** |
|
| 44 |
+
|-------------|-----------------|----------------|
|
| 45 |
+
| **Text** | 210 (46.6%) | English words encoded through temporal noise patterns |
|
| 46 |
+
| **Object Images** | 156 (34.6%) | Single objects encoded using temporal animation |
|
| 47 |
+
| **Dynamic Scenes** | 57 (12.6%) | Video depth maps with temporal motion patterns |
|
| 48 |
+
| **Shapes** | 28 (6.2%) | Geometric patterns encoded through temporal sequences |
|
| 49 |
+
| **Total** | **451** | **Comprehensive temporal understanding evaluation** |
|
| 50 |
+
|
| 51 |
+
**Download:** You can download the dataset from Hugging Face using the following command:
|
| 52 |
+
|
| 53 |
+
```bash
|
| 54 |
+
wget https://huggingface.co/datasets/timeblindness/spooky-bench/resolve/main/spooky_bench.zip
|
| 55 |
+
unzip spooky_bench.zip
|
| 56 |
+
```
|
| 57 |
+
|
| 58 |
+
**License:** MIT License
|