the details are given in publication [1,2,3]
File List:
MOS Folder: This Folder contains various MOS Score files. These files are: FileNameWithMOS.csv: contains the name of audiovisual files and MOS scores computed. BaseDataset.csv: Contains 5 independen parameters (video frame rate, quantization parameter, noise reduction, packet loss rate for audio and video streams) and MOS scores computed [1, 2]. WekaBaseDataset.arff: Same as BaseDataset in Weka format [1, 2]. ParametricDataset.csv: contains BaseDataset and other dependent parameters extracted from transmitted videos. Models using this dataset would be called parametric No-Reference models [1, 2]. WekaParametricDataset.arff: Same as ParametricDataset in Weka format [1, 2]. BitstreamDataset.csv: contains BaseDataset and other dependent parameters extracted from transmitetd videos as well as reduced amount of information from original videos. Models using this dataset would be called bitstream reduced-Reference models [3].
OutputVideoFiles Folder: This Folder contains saved output video files, RTCP stats collected during streaming for each file, consolidated statistics and video file header parameters, and randomized file list used in https://github.com/edipdemirbilek/SubjectiveAssesmentVideoPlayer.
SourceVideoFiles Folder: It includes source video files, and audio and video CAPS used in GStreaming.
SubjectDetails Folder: It contains ACR scores collected from all observers. These are non processed raw scores. No consolidation or rejection criteria is applied.
SubjectMerged Folder: Consolidated ACR scores with summary info. The scores here are after rejection criteria is applied. Refer to [1] for details.
CalculateCorrelation.py: Calculates the correlation of scrores of a given Subject with the overall MOS of all Subjects.
GenerateGraphs.py: Generates variosu graphs used in [1]
MergeSubjectDetails.py: Consolidates the ACR scores for a given Subject and also applies the rejection criteria specified in [1]
MergeSubjects.py: Consolidate all Subject scores and compines with detailed parameters and generates varios MOS files. Comments in the file itself.
PrepareBitstreamStats.py: From GStreamer RTCP stats file, it extracts only selected fields and stores in csv file.
SubjectsSummary.csv: includes, total number of accepted scores, average time to rate and correlation to the rest for each subject.
Publications related to this dataset:
For the Base and Parametric version of the dataset refer to:
[1] Demirbilek, Edip, and Jean-Charles Grégoire. “The INRS Audiovisual Quality Dataset." 2016 ACM Multimedia Conference.
[2] Edip Demirbilek and Jean-Charles Grégoire. Machine learning based parametric audiovisual quality prediction models for realtime communications. ACM TOMM (Revised).
For the Bitstream version of the dataset refer to:
[3] Edip Demirbilek and Jean-Charles Grégoire. Machine learning based bitstream audiovisual quality prediction models for realtime communications. IEEE International Conference on Multimedia and Expo, 2016 (Submitted).
For the implementation of the testbed used to create this dataset refer to:
[4] https://github.com/edipdemirbilek/GStreamerMultimediaQualityTestbed
[5] Edip Demirbilek and Jean-Charles Grégoire. Multimedia communication quality assessment testbeds. arXiv preprint arXiv:1609.06612, (2016).