P.1203/VQM: Qualitative Metrics for HTTP Adaptive Streaming

P.1203/VQM: Qualitative Metrics for HTTP Adaptive Streaming - dataset and software

The following outlines the various files and scripts to use with our P.1203 and VQM datasets and software.

Details on the P.1203 JSON files with reference to our original DASH datasets:

  1. P.1203 Dataset - The following tables provide links to the compressed P.1203 JSON input files.
  2. Files are compressed per P.1203 mode, with reference to our original DASH dataset.
  3. Full High Definition (Full HD - 1920x1080) is available in both copyright and non-copyright content.
  4. The copyright content, 20 clips, is typically used for trace-based simulation, while the non-copyright content, 3 clips, providing metrics for decodable streamed content.
  5. Ultra High Definition (UHD - 3840x2160) content, 3 clips, is only available for P.1203 up to the Full HD resolution and is non-copyright content.
  6. Each zip file contains a P.1203 JSON file per segment, for each clip and across five different segment durations (2-sec, 4-sec, 6-sec, 8-sec, and 10-sec).
  7. Content duration is typically between 10 minutes and 16 minutes in duration.


FULL HD Copyright Content:

P.1203 P.1203 P.1203 P.1203
mode 0 mode 1 mode 2 mode 3


FULL HD Non-Copyright Content:

P.1203 P.1203 P.1203 P.1203
mode 0 mode 1 mode 2 mode 3


UHD Copyright Content:

P.1203 P.1203 P.1203 P.1203
mode 0 mode 1 mode 2 mode 3

The branched version of dashc can be found here. Additional options in this version of dashc include:

  1. [-extended], display aditional columns with information on QoE metrics - for P.1203 this will equate to 'true'
  2. [-preferred-codec], (h264, h265, vp09, av01) stream using the specified codec - for P.1203 this will equate to 'h264'
  3. [-max-rep-height] the maximum resolution height to stream at (defauls as 2160) - for P.1203 this will equate to '1080'

The VQM dataset is generated using the Command Line Video Quality Metric \textit{(CVQM)} tool available here, which is written in MATLAB. CVQM was ported to Ubuntu OS and Python script are developed to automate the calculation of VQM values for individual segments of our Full HD video dataset. The Python script, available here, while the generated dataset is available here. As the only limitation in VQM is a sufficiently large segment duration, no less than 4 seconds, we were also able to generate content for the H.265 (HEVC) element of the Full HD dataset.

A VirtualBox VM of Ubuntu 18.04, containing all the open-source programs needed to stream our UHD and Full HD Datasets using the extended version of our dashc player is available here. The extended version of dashc will generate additional columns including a P.1203 video QoE metric per row (per segment request).

Sample output of dashc is shown below complete with summary details:

WARNING: WidthxHeigth (3840x2160) of clip @ rate 41.66Mbps - is greater than defined max_rep height 1080
WARNING: WidthxHeigth (3840x2160) of clip @ rate 25.99Mbps - is greater than defined max_rep height 1080
WARNING: WidthxHeigth (3840x2160) of clip @ rate 15.56Mbps - is greater than defined max_rep height 1080
 Seg_#  Arr_time  Del_Time  Stall_Dur  Rep_Level  Del_Rate  Act_Rate  Byte_Size  Buff_Level   RTT   Codec  Width  Height  FPS  Seg_Dur  Start  P.1203 QoE
 1    30     30     0.000000    239    551    8     2132    2.000    0.061644  h264   320   180   24   2.000   0.000  1.936882
 2    1248      555    0.000000    378    1970     547    136969   4.000    0.061612  h264   384   216   24   2.000   2.000  1.936882
 3    5423    3512    0.000000    378    410    720    180164   2.488    0.061771  h264   384   216   24   2.000   4.000  1.939419
 4    7163    1078    0.000000    239    802    432    108123   3.409    1.118332  h264   320   180   24   2.000   6.000  1.938224
...
...
 147   187992   429      0.000000    239    1287     276    69090   209.526     0.061644  h264   320   180   24   2.000  292.000  1.951534
 148   188963   308      0.000000    239    1858     286    71640   211.218     0.061690  h264   320   180   24   2.000  294.000  1.951534
 149   189712   87        0.000000    239    4477     194    48706   213.131     0.061546  h264   320   180   24   2.000  296.000  1.951534
 150   190492   118      0.000000    239    2497     148    37007   215.012     0.061612  h264   320   180   24   2.000  298.000  1.951534

Downloaded 150 segments (2-second h264) totalling  9.31 MB

P.1203 Quality of Experience:       1.9515
Round Trip Time (RTT) min/avg/max:    0.061/0.103/1.486
Delivery rate min/avg/max:        199.453/1460.620/6683.705 KB/s
Number of stalls/Total stall time:      0/0.000
Representation level min/avg/max:      239.515/241.369/378.577 KB/s
Resolution min/mode/max:         320x180/320x180/384x216

Enjoy, Jason (j.quinlan@cs.ucc.ie)

This work has emanated from research conducted with the financial support of Science Foundation Ireland (SFI) under Grant Number 13/IA/1892.

 

Mobile and Internet Systems Laboratory

Department of Computer Science, Western Gateway Building, University College Cork, Western Road, Cork, Ireland.

Top