Data Preparation

[yoohj0416]

This issue is a set of steps for data preparation.

• [ ] Donwload all videos (1.8TB)
• [x] Make a code which extracts frames from video with FFmpeg
• [ ] Extract frames from video
• [x] Organize tag information from BDD100K labels (i.e. weather, scene, and others)

[yoohj0416]

BDD100K labels include following items:

names
attributes
  └ weather: clear, overcast, snowy, rainy, cloudy, foggy
  └ scene: city street, tunnel, highway, residential, parking, gas station
  └ timeofday: daytime, dawn/dusk, night
timestamp
labels
  └ instance: Car, Sign, Light, Person, Truck, Bus, Bike, Rider, Motor, Train (Occluded or Not, Truncated or Not)
  └ lane: road curb, crosswalk, double white, double yellow, single white, single yellow (full or dashed, parallel or perpendicular)
  └ drivable area: directly drivable or alternatively drivable

Overview of BDD100K