Wyniki wyszukiwania - BazTech

Ograniczanie wyników

Znaleziono wyników: 2

Liczba wyników na stronie

Wyniki wyszukiwania

Wyszukiwano:
w słowach kluczowych: Netflix

Sortuj według:

Ogranicz wyniki do:

Subjective quality evaluation of 8- and 10-bit MP4-coded video sequences from Netflix

Falkowski-Gilski Przemysław, Uhl Tadeus, Divakarachari Parameshachari Bidere

Zeszyty Naukowe Politechniki Morskiej w Szczecinie

2024

nr 79 (151)

30--38

Recently, many researchers have been intensively conducting quality of service (QoS), quality of experience (QoE), and user experience (UX) studies in the field of video analysis. This paper is intended to make a new, complementary contribution to this field. Currently, streaming platforms are key products in relation to delivering video content online. Most often, they include the MP4 video format, which is most widely utilized among audio-visual codecs. This study involves a group of 38 individuals, aged between 21–35 years old, in a laboratory consisting of 20 iMacs with 4K retina display. The presented signal sequences included content sourced from the Netflix Chimera repository, with 8- and 10-bit depth, available in different resolutions of 270p, 432p, 720p, and 1080p. Tests included a subjective quality evaluation in a 5-step mean opinion score (MOS) scale, focused on the UX aspect. According to the obtained results, content with the lowest and highest resolutions is optimal in 8-bit depth, while movies with intermediate resolutions are better in 10-bit depth. For 8-bit content, the main problem is pixelation, whereas, in the case of 10-bit samples, the main issue is color noise, particularly in the case of the lowest resolution. Many viewers indicated that 10-bit encoding offered lower quality. Moreover, 8-bit movies caused a lower quality of the gradient, presumably due to the smaller range of the available color. However, 8-bit movies in the same situation generate visible stripes on static images in the background, causing a lower quality of the gradient, which is probably due to the smaller range of available colors. The results of the performed experiments may be of particular interest to content creators and distributors, particularly network and cable operators, as well as wireless and wired providers. coding, compression, QoE (Quality of Experience), UX (User Experience), video content, Netflix.

Analyzing certain temporal dependences in Netflix data

Łupińsla-Dubicka A., Drużdżel M.

Zeszyty Naukowe Politechniki Białostockiej. Informatyka

2008

Z. 3

67-82

Netflix (see http://www.netflix.com/), an American Internet-based movie rental company, uses data mining in their recommendation system. In October 2006 Netflix made a huge data base of their users and movie evaluations available to the community and announced a million dollars prize to the team that beats the accuracy of their recommendations by at least 10%. The data have since become an object of interest of the machine learning community. In this paper, we focus on one aspect of the data that, to our knowledge, has been overlooked — their temporal dependences. We have looked at the impact of the day of the week, month of the year, length of membership, month from the start of Netflix, etc., on the average evaluation.

Działający w Stanach Zjednoczonych Ameryki Netflix (http://www.netflix. com/) jest jedną z największych na świecie internetowych wypożyczalni filmów. W celu uzyskania wyższej jakości proponowanych przez system ocen filmów, w październiku 2006 roku Netflix udostępnił bazę danych użytkowników oraz ich ocen i ogłosił nagrodę dla tego, kto uzyska co najmniej 10-cio procentową poprawęw stosunku do wyników Cinematch (RMSE=0.9525). W tym artykule postawiliśmy sobie za cel zbadanie, czy zalezności czasowe, takie jak dzień tygodnia lub długość członkostwa, są w stanie zwiększyć jakość prognozowania.