# VQA - AUTOPILOT CVPR — Data

来源：https://www.kaggle.com/competitions/AUTOPILOT-VQA/data

本地 Meta Kaggle 快照；不是本次实时页面。

# Dataset Description

This page describes the files provided for the AUTOPULOT VQA challenge and the exact format required for submission. The task is multi-output classification: for each video you must predict one integer label per question column, using the legend below.​

##Data sources and attribution

The 2COOOL benchmark builds on prior 2COOOL/COOOL efforts and integrates clips originating from multiple dashcam datasets, including COOOL, DADA-2000, and Nexar collision footage.
If you use this dataset in your research (ie releasing a new dataset based on these videos), please consider citing the underlying sources and follow the attribution/redistribution requirements in the competition “Rules” page.

**While we suggest citing the works that form this challenge dataset, participants are not required to do so as a condition of using a dataset or participating in the challenge or submitting a workshop paper. Though please be mindful of the specific licensing terms of the Nexar dataset.**

##Files:

sample_submission.csv — A valid submission template with the correct header, column names, and ordering.​

Please find the dataset files (sample_submission, videos and heatmaps) on Google Drive: https://drive.google.com/drive/folders/1JYhSPv0zQQ_EBx2G4ljyF6CNXlxkRNVB


##What you predict:

For every video_id, predict integer-coded answers to the following question set:

1. Weather
2. Light Condition
3. Traffic Environment
4. Road Configuration
5. Incident Type (category, primary/secondary incident entities, vehicle types where applicable, behaviors, incident peak)
6. Incident Prevention Measure
7. Initial Point of Impact
8. Traffic Control & Signage Presence
9. Road Surface & Material Condition

Each question corresponds to one column in sample_submission.csv (besides video_id). The exact column names include line breaks and must match the provided files exactly.

##Submission format (CSV)

Your submission must be a CSV with:

- One row per video_id in the test set (NOTE: video 123.mp4 should just be the integer 123 in the video_id column)

- One column per question (same header and order as sample_submission.csv)

- Integer predictions only

- **Participants are highly advised to use column names copied directly from sample_submission.csv**

##Label legend (all answers are numeric)

Use the following mapping for each question column. If the correct answer is unknown, use -1.

###A) Weather

Q1.a: What was the time of day during the incident?
- Day = 0
- Dawn (sunrise) = 1
- Dusk (sunset) = 2
- Night = 3
- Unknown = -1

Q1.b: What were the weather conditions during the incident?
- Sunny/Clear = 0
- Cloudy = 1
- Rainy = 2
- Partly Cloudy = 3
- Snowy = 4
- Foggy = 5
- Unknown = -1

###B) Light Condition

Q2: What was the lighting condition at the time of the incident?
- Daylight = 0
- Headlights only = 1
- Sunrise/Sunset – Early morning low-light, sun beginning to rise = 2
- Streetlights on = 4
- Unknown = -1

###C) Traffic Environment

Q3.a: Where did the incident take place (traffic environment)?
- Urban area – City streets with dense buildings, intersections, and traffic = 0
- Suburban area – Residential or mixed-use commercial/residential area = 2
- Rural area – Open or undeveloped roadside with sparse traffic = 3
- Unknown = -1

Q3.b: What facilities were present in the surrounding environment?
- Urban area - skyscrapers, businesses, etc. = 1
- Construction zone / Work area – Road partially blocked or altered due to work = 2
- Residential Area = 3
- Parking lot / Gas station – Off-road traffic environment with vehicles maneuvering = 4
- Near a school/college – Educational zone with potential pedestrian activity = 5
- Unknown = -1

###D) Road Configuration

Q4.a: What is the road configuration at the incident time?
- Highway = 1
- Intersection = 2
- T-junction / Y-junction = 3
- Residential street = 4
- Bridge / Overpass / Underpass = 5
- Roundabout / Traffic circle = 6
- Exit/Entrance ramp = 7
- Unknown = -1

Q4.b: Road lane configuration
- Two-way divided = 0
- Two-way not divided = 1
- One-way = 2
- Unknown = -1

Q4.c: Lane direction of Ego-Car at time of incident
- Right = 0
- Left = 1
- Middle = 2
- Unknown = -1

###E) Incident Type

Q5.a: What is the category of the incident?
- Pedestrian/Cyclist/Vulnerable Road Users = 0
- Vehicle-to-Vehicle = 1
- Animals / Objects = 2
- No Collision = 3
- Barriers / Fixed Objects = 5
- Vehicle Loss of Control / Rollover = 6
- Unknown = -1

Q5.b: Primary Incident Entity (Who caused or could have prevented the incident)?
- Pedestrian = 0
- Ego-car = 1
- Animal = 2
- No collision = 3
- Another Vehicle = 5
- Cyclist = 6
- Flying Object = 10
- Smoke = 13
- Object on the road = 29
- Unknown = -1

Q5.c: If Primary Incident Entity is a vehicle, what type of vehicle is it?
- Small Car = 0
- No collision = 1
- Truck = 2
- Bicycle = 3
- Bicycle Cart = 4
- Motorcycle = 5
- Scooter = 6
- SUV = 8
- Bus = 9
- Unknown = -1

Q5.e: Primary Entity Behavior
- Crossing = 0
- Speeding = 1
- Driving normally = 3
- Overtaking/Changing Lanes = 4
- Turning = 6
- Walking = 7
- Stopped = 8
- Swerving = 10
- Ignoring traffic signal = 11
- Lost control = 12
- Fell Down = 13
- Rolling through = 14
- Failure to yield = 15
- Flying = 16
- Rolling over = 17
- Stationary = 18
- Fallen into the road = 19
- Unknown = -1

Q5.f: Secondary Incident Entity (Victim, secondary cause or entity that was hit)
- Cyclist = 0
- Ego-car = 1
- Another Vehicle = 2
- Object or Barrier = 3
- Pedestrian = 4
- Scooter = 5
- Animal = 6
- Unknown = -1

Q5.g: If Secondary Incident Entity is a vehicle, what type of vehicle is it?
- Scooter = 0
- Small Car = 1
- Bicycle = 2
- Truck = 3
- SUV = 4
- Bus = 5
- Unknown = -1

Q5.i: Secondary Entity Behavior
- Driving normally = 0
- Swerving = 1
- Fixed Position (also for stationary cars) = 2
- Speeding = 3
- Crossing = 4
- Turning = 5
- Lost control = 6
- Failure to yield = 7
- Flying = 8
- Stopped = 9
- Fallen onto road = 10
- Overtaking/Changing Lanes = 11
- Rolling over = 13
- UTurn = 14
- Unknown = -1

Q5.j: Incident peak
- Near-miss = 0
- No collision = 1
- Collision = 2
- No effect = 3
- Avoided in advance = 4
- Chain-reaction = 5
- Unknown = -1

###F) Incident Prevention Measure

Q6.a: How could this Incident most likely have been prevented by the Primary Entity?
- Primary Entity was traveling unsafe = 0
- No Fault: Primary Entity was behaving normally but could not prevent incident = 1
- Primary Entity could have maneuvered but did not = 2
- No Incident = 3
- No Fault: Primary Entity avoided Incident = 4
- Primary Entity is not capable of preventing accident (animal/inanimate) = 5
- Primary entity had defect (object/part fell off, equipment failure) = 6
- Unknown = -1

Q6.b: How could this incident most likely have been prevented by the Secondary Entity?
- No Fault: Secondary Entity avoided Incident = 0
- No Fault: Secondary Entity was behaving normally but could not prevent accident = 1
- Secondary Entity is not capable of preventing accident (animal/inanimate) = 2
- Secondary Entity was traveling unsafe = 3
- No Incident = 4
- Secondary entity had defect (object/part fell off, equipment failure) = 5
- Secondary Entity could have maneuvered but did not = 6
- Unknown = -1

###G) Initial Point of Impact

Q7.a: If the Primary Entity was a vehicle, where on the vehicle did the initial impact occur?
- No collision = 0
- Front = 1
- Left side = 2
- Right side = 4
- Rear = 5
- Top/Roof = 6
- Unknown = -1

Q7.b: If the Primary Entity was a vehicle, what side of the impacted area was hit?
- No collision = 0
- Left Corner = 2
- Front = 3
- Right Side of Vehicle = 4
- Right Corner = 5
- Undercarriage / Bottom = 6
- Left Side of Vehicle= 7
- Unknown = -1

Q7.c: If the Secondary Entity was a vehicle, where on the vehicle did the initial impact occur?
- No Collision = 0
- Right side = 1
- Front = 2
- Top/Roof = 3
- Left side = 5
- Rear = 6
- Unknown = -1

Q7.d: If the Secondary Entity was a vehicle, what side of the impacted area was hit?
- No collision = 0
- Left Corner = 1
- Right Corner = 3
- Left Side of the Vehicle = 4
- Roof/Top = 5
- Ride Side of Vehicle = 6
- Front Grill/Bumper = 7
- Mid-Right Side = 9
- Mid-Left Side of Truck = 10
- Undercarriage / Bottom = 11
- Rear = 12
- Unknown = -1

###H) Traffic Control & Signage Presence

Q8: What traffic control or road sign is present at the incident location?
- Yield / Give way sign = 0
- None visible = 1
- Traffic light = 3
- Construction / Work zone sign = 4
- No entry / One-way sign = 5
- Pedestrian crossing = 6
- Traffic light Ahead Sign = 7
- School zone / Warning sign = 8
- Speed limit sign = 9
- Stop sign = 10
- Railroad crossing sign / gate = 11
- Unknown = -1

###I) Road Surface & Material Condition

Q9.a: What was the road surface condition during the incident?
- Dry pavement = 0
- Wet road = 1
- Snow / Ice covered = 2
- Debris / Mud on road = 3
- Unknown = -1

Q9.b: What type of road material is the vehicle driving on?
- Asphalt / Blacktop = 0
- Concrete = 1
- Gravel / Dirt = 3
- Unknown = -1