Blurred Lines Unrated Version 1080p Download
Pablo Barjavel
The music video for "Blurred Lines" was directed by Diane Martel. Two versions of the video exist: edited and unrated. In both of them, Thicke, T.I., and Williams are featured with models Emily Ratajkowski, Elle Evans, and Jessi M'Bengue performing several activities, including the models snuggling in bed with Thicke and sitting on a stuffed dog. After being on the site for just under one week, the unrated version, featuring topless models, was removed from YouTube for violating the site's terms of service. Many critics panned both videos, calling them misogynist and sexist.
In an interview with GQ's Stelios Phili, Thicke explained that he and Williams were in the studio together when he told Williams that one of his favorite songs of all time was Marvin Gaye's 1977 single "Got to Give It Up". Thicke wanted to make a song similar to "Got to Give It Up". Thicke stated that he and Wiliams would go back and forth and sing lines like, "Hey, hey, hey!".[6] Thicke told the Daily Star the song was "mostly throwaway fun", but said it was inspired by him and Williams being in love with their wives, having kids, and loving their mothers. He commented that both of them have a lot of respect for women.[7] An ad was created for Radio Shack to market the Beats Pill, a small stereo, that showed Thicke, Pharrell, and the models repeating the look of the (clothed) music video, but with the models holding up the Beats Pill.[8]
"Blurred Lines" debuted at number 94 on the US Billboard Hot 100.[53] After the song's unrated version of the video was released, the song rose from number 54 to number 11.[54] The track rose from number 11 to number 6, giving Thicke his first top 10 hit in the US.[55] The song would later rise from number six to number one in June 2013, giving T.I. his fourth, Pharrell his third, and Thicke's first number one hit in the US.[56] "Blurred Lines" topped the Hot 100 for 12 consecutive weeks, making it the longest running single of 2013.[57][58] Billboard named "Blurred Lines" the song of the summer in September 2013.[59] On the Billboard Hot R&B/Hip-Hop Songs chart, the song reigned at number one for 16 weeks, making it one longest tracks to stay at number one on the chart.[60] In June 2018, The single was certified a diamond certification by the Recording Industry Association of America (RIAA), denoting track-equivalent sales of 10,000,000 units in the US based on sales and streams.[61]
A music video for "Blurred Lines" was directed by Diane Martel and was released on March 20, 2013,[72] while an unrated version was released on March 28, 2013.[73] After being on the site for just under one week, the unrated version of the video was removed from YouTube on March 30, 2013, citing violations of the site's terms of service that restricts the uploading of videos containing nudity, particularly if used in a sexual context.[74][75] However, it was later restored on July 12, 2013.[76] The unrated video remains available on Vevo, while the edited version is available on both Vevo and YouTube.[77][78][79] The unrated version of "Blurred Lines" generated more than one million views in the days following its release on Vevo.[80] Thicke told GQ they wanted to do "old men dances" and imitate how they were in the studio. They tried to do everything that was prohibited by social custom. He stated they did bestiality, drug injections, and things that are derogatory towards women. When it came to the balloon arrangement, Thicke said it was Martel's idea. They wanted to "go over the top" and be as witless as possible.[6]
A basic question to ask when applying a new set of features to the task of categorical search is whether different features are needed to accomplish the two subtasks of guidance and recognition; are the features used to guide gaze to a categorically defined target the same as the features used to recognize that object as a target once it is fixated? As already noted, the visual search community has not vigorously engaged this question, and in fact has seemed content with the assumption that search guidance and object recognition use different features that are tailored to the specific demands of the different tasks. There is even good reason to suspect why this might be true. By definition, the features used to guide gaze to an object must work on a blurred view of that object as it would be perceived in peripheral vision. The features used for recognition, however, would be expected to exploit high-resolution information about an object when it is available. Color is another example of a potentially asymmetric use of object information. Search guidance has long been known to use color (Rutishauser & Koch, 2007; Williams, 1967; Williams & Reingold, 2001; see also Hwang, Higgins, & Pomplun, 2007, for guidance by color in realistic scenes), presumably because of its relative immunity to the effects of blurring. Object recognition, however, places less importance on the role of color (Biederman & Ju, 1988), with many behavioral and computer vision models of recognition ignoring color altogether (e.g., Edelman & Duvdevani-Bar, 1997; Hummel & Biederman, 1992; Lowe, 2004; Riesenhuber & Poggio, 1999; Torralba, Murphy, Freeman, & Rubin, 2003; Ullman, Vidal-Naquet, & Sali, 2002).
Representative search displays used in the behavioral experiment, illustrating the relationship between object eccentricity and retinal acuity. (A) Target present display, with the teddy bear target shown enlarged at inset. Note that all objects would be perceived as blurred when viewed from a central starting fixation position (blue dot). (B) The same target in the same search display viewed after its fixation; the red arrows and blue dots show representative eye movements and fixations during search. (C) A target-absent trial in which the first fixated object was a high-similarity bearlike distractor, again with representative search behavior.
Figure 3B replots the behavioral data with the corresponding data for the nine models tested, each color coded and plotted in order of decreasing match to the human behavior. This ordering of the data means that the best matching model is indicated by the leftmost colored bar for each object type. As in the case of the behavioral data, each colored bar indicates the probability that a given object detector would have selected a particular type of object for immediate fixation. This selection was made on a trial-by-trial basis, as described in the computational methods section; the likelihood of an object being a bear was obtained for each of the four objects in a search display, with our prediction of the first-fixated object on that trial being the one with the highest likelihood estimate. This again was done on blurred versions of each object, so as to approximate the visual conditions existing at the time of search guidance.
To explicitly evaluate the assumption that search guidance and recognition are separate processes we explored two classes of models, one approximating the conditions existing during guidance and the other approximating the conditions existing during recognition. The guidance models were given sets of four blurred objects, each corresponding to the objects in a search display, and predicted the object that would be fixated first. The recognition models were given unblurred versions of these same first-fixated objects, and classified these objects as teddy bear targets or nonbear distractors. We also manipulated the visual similarity between these objects and the target class, as well as the types of features and/or methods that were used by the models. In the context of this categorical bear search task, we found that guidance and recognition could be well described by several relatively simple computational models, all without the use of any explicit fit parameters. However, of the nine models that we tested the one that best predicted both categorical guidance and recognition was not an implausibly complex model from computer vision, but rather a basic version of an HMAX model, one that included a color feature. This HMAX+COLOR model not only captured the finding that gaze was guided to nontarget objects in proportion to their similarity to the target class, it also captured the magnitude of these guidance effects for each of the similarity conditions that we tested. This same model also captured the behavioral false negative and false positive rates, as well as the effect of target-distractor similarity on these recognition errors. In summary, under conditions that closely approximate the information available to observers, namely whether the objects were blurred or not depending on viewing from pre- or post-fixation, we found that the HMAX+COLOR model was able to predict behavioral guidance and recognition during a categorical teddy bear search with impressive accuracy.
However, our suggestion that guidance is a form of preliminary recognition performed on blurred objects does not require the full preattentive recognition of every object in a search display. The reason for this stems from the distinction between object recognition and object detection. Recognition is the attachment of meaning to a pattern via comparison to patterns that have been learned and committed to memory. This is true for both biological and computer systems. Preattentive recognition would mean that this process occurs automatically and in parallel for every pattern appearing in a scene. Walking into an opening reception of a conference would therefore cause names to be attached to all familiar attendees, and every other object in the scene, regardless of where or how attention was allocated. Detection is the determination of whether a scene contains a particular pattern, with preattentive detection being the simultaneous comparison of this target pattern to all the patterns appearing in a scene. The analogous preattentive detection task would be walking into an opening reception with the goal of finding a particular colleague, and having all the patterns in a scene automatically evaluated and prioritized with respect to this goal. The product of a preattentive detection process is therefore a target map (Zelinsky, 2008) or a priority map (Bisley & Goldberg, 2010) that can be used to guide overt or covert attention; the product of a preattentive recognition process would be a map of meaningful objects, with the semantic properties of each being available to guide attention. Although building a target map is not trivial, especially when targets can be entire object classes, this task is vastly simpler than the task of building a map of recognized objects. This latter possibility has even been criticized as being biologically implausible on the basis of computational complexity, with the comparison of every pattern in a scene to every pattern in memory potentially resulting in a combinatorial explosion (Tsotsos, 1990).
e2b47a7662