=====================================================
SLICE - Street-Level Insights from Camera Evidence
Authors: Peter Roch, Lukas Laaser, Bijan Shahbaz Nejad, Marcus Handte, and Pedro José Marrón
Affiliation: University of Duisburg-Essen, Essen, Germany
Contact: peter.roch@uni-due.de

-----------------------------------------------------
1. Overview
-----------------------------------------------------
The SLICE dataset provides a large-scale collection of traffic video recordings from urban surveillance cameras in California.  
It is intended for research in traffic monitoring, vehicle detection, and congestion analysis using computer vision.
The dataset is available at https://www.nes.uni-due.de/wp-content/data/slice/

-----------------------------------------------------
2. Dataset Contents
-----------------------------------------------------
- README.txt				This README file
- downloader.py				a python script assisting with the download of the dataset

- metadata/					metadata / categories - Each directory contains txt files listing all videos in this category
- metadata/bitrate/			videos split into quartiles by bit rate
- metadata/duration/		videos split by duration (short, medium short, medium long, long)
- metadata/fps/				videos split by fps (15, 30, low, medium, high)
- metadata/frame_count/		videos split into quartiles by frame count
- metadata/resolution/		videos split by resolution (320x240, 352x240, 640x480, 704x480, 720x480, 768x432, 1280x720, 1920x1080)
- metadata/times/			videos split by recording time (dawn, morning, day, evening, dusk, night)
- metadata/traffic/			videos split by traffic volume (light, moderate, high, dense, peak
- metadata/weekdays/		videos split by weekday (monday, tuesday, wednesday, thursday, friday, saturday, sunday)
- metadata/video_info.csv	a csv file listing technical properties of each video

- recordings/				all recordings, split into directories of {city}/{camera}/{day}/{video}.mp4
							Additionally, each camera directory contains a file "camera_info.txt" containing information about the camera at the time of recording

							city: 	one of Los Angeles, Sacramento, San Francisco
							camera: the name of the camera, according to Caltrans API
							day: 	the day in the format YYYYmmdd (i.e. year, month, day)
							video:	the filename has the format YYYYmmddHHMMSS------0700 (i.e. year, month, day, hour, minute, second, microsecond, timezone of california)
							Example: recordings/Sacramento/Hwy 5 At Arena Blvd 1/20250404/202504040722395550410700.mp4 
									 - This video was recorded in Sacramento using the camera "Hwy 5 At Arena Blvd 1" on April 4, 2025, at 07:22 a.m.
							
-----------------------------------------------------
4. Download Instructions
-----------------------------------------------------
Using the provided downloader.py script, you can download the complete dataset or different parts.
The script has been tested with Python 3.10, but should run on all supported versions of Python.
The script uses tqdm and requests. You can install those libraries with "pip install tqdm requests".
The help instructions can be listed as follows:
$ python downloader.py --help
usage: downloader.py [-h] [-o OUTPUT] [-all] [-ci] [-m] [-br {Q1,Q2,Q3,Q4}] [-d {s,ms,ml,l}]
                     [-fps {15,30,low,medium,high}] [-fc {Q1,Q2,Q3,Q4}]
                     [-res {320x240,352x240,640x480,704x480,720x480,768x432,1280x720,1920x1080}]
                     [-t {dawn,morning,day,evening,dusk,night}]
                     [-wd {monday,tuesday,wednesday,thursday,friday,saturday,sunday}]
                     [-tr {light,moderate,high,dense,peak}] [-v VIDEO]

options:
  -h, --help            show this help message and exit
  -o OUTPUT, --output OUTPUT
                        output directory
  -all, --all           download the complete dataset
  -ci, --camera_info    download camera info txt files
  -m, --metadata        download csv files with metadata
  -br {Q1,Q2,Q3,Q4}, --bitrate {Q1,Q2,Q3,Q4}
                        download videos with bitrate in the specified quartile
  -d {s,ms,ml,l}, --duration {s,ms,ml,l}
                        download videos with specified duration
  -fps {15,30,low,medium,high}, --fps {15,30,low,medium,high}
                        download videos with specified fps
  -fc {Q1,Q2,Q3,Q4}, --frame-count {Q1,Q2,Q3,Q4}
                        download videos with frame count in the specified quartile
  -res {320x240,352x240,640x480,704x480,720x480,768x432,1280x720,1920x1080}, --resolution {320x240,352x240,640x480,704x480,720x480,768x432,1280x720,1920x1080}
                        download videos with specified resolution
  -t {dawn,morning,day,evening,dusk,night}, --time {dawn,morning,day,evening,dusk,night}
                        download videos during the specified time
  -wd {monday,tuesday,wednesday,thursday,friday,saturday,sunday}, --weekday {monday,tuesday,wednesday,thursday,friday,saturday,sunday}
                        download videos from the specified weekday
  -tr {light,moderate,high,dense,peak}, --traffic {light,moderate,high,dense,peak}
                        download videos with the specified traffic volume
  -v VIDEO, --video VIDEO
                        download the specified video

-----------------------------------------------------
5. License
-----------------------------------------------------
The SLICE dataset is distributed under the Creative Commons Attribution 4.0 International Public License (CC BY 4.0) - https://creativecommons.org/licenses/by/4.0/.  
Please cite our paper when using this dataset in your research.

-----------------------------------------------------
6. Citation
-----------------------------------------------------
Peter Roch, Lukas Laaser, Bijan Shahbaz Nejad, Marcus Handte, Pedro José Marrón: SLICE - Street-Level Insights from Camera Evidence. In: Advances in Visual Computing, Forthcoming.

=====================================================