File size: 3,557 Bytes
5e6d39f a4a9519 5e6d39f a4a9519 62e01a9 5e6d39f | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 | <div align="center">
<a href="https://github.com/SafeRL-Lab/Open-Space-Reasoning">
<img src="./figures/logo-full.png" alt="Logo" width="60%">
</a>
<h1 align="center" style="font-size: 30px;"><strong><em>AccidentBench</em></strong>: Benchmarking Multimodal Understanding and Reasoning in Vehicle Accidents and Beyond</h1>
</div>
### Dataset Format:
```jsonc
{
"id": int,
"dataset": "str", // e.g., sub dataset filename
"scene_name": "str", // e.g., video filename
"reasoning_style": "str", // e.g., temporal_reasoning, intent_goal_reasoning, etc.
"question": "str", // The reasoning question related to the scene
"ground_truth": "str", // Correct answer key (e.g., "A", "B", etc.)
"options": ["str", "str", "str", "str", "str", "str"] // Multiple-choice options
}
```
One example from air space:
```jsonc
{
"id": 1,
"dataset": "air_space_long",
"scene_name": "air_space_long_1.mp4",
"reasoning_style": "intent_goal_reasoning",
"question": "How many moving airplanes are observed in this video?",
"ground_truth": "A",
"options": [
"E. [0,1]",
"C. [8,9]",
"A. [4,5]",
"D. [6,7]",
"B. [10,11]",
"F. [2,3]"
]
}
```
<div align=center>
<img src="./figures/qa-example.png" width="85%"/>
</div>
<div align=center>
<center style="color:#000000;text-decoration:underline">Figure 1. A question and answer example: For each open-space reasoning setting, we include three
types of video lengths: short, medium, and long. Each video length includes tasks designed to
evaluate temporal reasoning, spatial reasoning, and intent reasoning.</center>
</div>
### Dataset Distribution:
<div align=center>
<img src="./figures/data_distribution.png" width="85%"/>
</div>
<div align=center>
<center style="color:#000000;text-decoration:underline">Figure 2. Distribution of video and task properties in the M4R benchmark.</center>
</div>
### Three Space Settings
<div align=center>
<img src="./figures/three-example-scenarios.png" width="85%"/>
</div>
<div align=center>
<center style="color:#000000;text-decoration:underline">Figure 3. Examples of multimodal Understanding and Reasoning in Open-Space Scenarios.</center>
</div>
### Reasoning Settings:
<div align=center>
<img src="./figures/reasoning-settings.png" width="85%"/>
</div>
<div align=center>
<center style="color:#000000;text-decoration:underline">Figure 4. Examples of reasoning question settings in M4R across three key reasoning types: Temporal
Reasoning, which involves understanding event sequences and motion over time; Spatial Reasoning,
which focuses on relative positioning and orientation in space; and Intent Reasoning, which evaluates
understanding of goal-directed behaviors and decision-making in dynamic environments..</center>
</div>
### One Example in Land Space Settings:
<div align=center>
<img src="./figures/land-space-examples.png" width="85%"/>
</div>
<div align=center>
<center style="color:#000000;text-decoration:underline">Figure 5. Land-space traffic accident scenarios for open-space video understanding and reasoning include
<span style="color:cyan;">intersection collisions</span>,
<span style="color:blue;">urban road accidents</span>,
<span style="color:gray;">nighttime incidents</span>,
<span style="color:orange;">rural road accidents</span>,
<span style="color:pink;">snow-covered road collisions</span>, and
<span style="color:green;">freeway accidents</span>.</center>
</div>
|