Home / Fqa's / frequently asked  questions


Advances in both Behavioral Vision Science and Computational Vision are the basis for computational models of human visual attention that take as input an image or video and make predictions about where people will fixate in the first 3 to 5 seconds of viewing. Initial fixations play an important role in acquiring an understanding of a scene and/or content and serve as the gateway to further visual analysis. 3M Visual Attention Service (3M VAS) uses a computational model of visual attention to make predictions about where initial fixations will occur in an image. Validation studies are presented in which eye-tracking results are compared to the predictions made by 3M VAS on the same set of images. Using signal detection Response-Operator-Characteristics (ROC) we found that the model was able to predict human fixations at around 85% of the theoretical limit (theoretical limit being eye-tracking). If biases are removed from the eye-tracking data (e.g., the central bias which is a function of collecting data on computer monitors), the model performs at approximately 90% of the theoretical limit.


3M Visual Attention Service

Validation Study

3M Visual Attention Service is a software tool that takes as input an image or video and as output makes predictions about where people will initially look within that scene. The software is based on 30 years of research in academic institutions and recent research at 3M to better understand what people will initially notice. The goal of this research is to develop a deep understanding of how visual attention is initially allocated in a complex scene—that is where will people look—and to translate that knowledge to a software tool that predicts where these initial fixations will occur.


Performance Evaluation

  • 1. Introduction

    There has been a great deal of research in Vision Science and

    Computer Vision to develop theories of visual attention that

    are converted into mathematical equations and algorithms

    that make explicit predictions about where people will

    initially look when viewing complex scenes such as

    shopping malls, streets, magazine pages, web pages,

    advertising content, etc.


    But how does the visual system make decisions on where

    to look? The human visual system takes in a large amount

    of visual information (that is, the entire visual field) at any

    given time. However, the human visual system expends the

    bulk of its resources processing only a small fraction of this

    information—typically where the viewer is fixated. Although

    the majority of the visual field is not being attended to, it

    serves as an important part of the visual experience. The

    human visual system uses the information in the periphery to

    monitor regions that might be of interest to the viewer—the

    regions that attract visual attention. If the early perceptual

    properties (color, motion, contrast etc.) are engaging, the

    human visual system will move its fixation to that location

    to gather more visual information. Thus, when one initially

    enters a room, or initially turns a corner, or looks at a

    magazine page, the human visual system surveys the scene to

    become aware of what might be important within the scene.

    Based on prior research, during this initial “surveillance”

    period our visual system is attracted to low-level perceptual

    features such as color, luminance, edges, motion and other

    early visual processing elements. This initial surveying

    typically occurs in the first 3-5 seconds. During this survey

    period the visual system analyzes the visual representation

    projected on the retina and based on that information the

    eyes are drawn to particular regions of the scene based on

    the visual features (e.g., color, contrast, text faces, etc.).

    What people do after this initial 3-5 seconds will depend

    on Top-Down processing. These secondary fixations that

    are primarily driven by Top-Down processing will use the

    information gathered during these initial fixations to decide




  • 2. Data & Methods

    L3M VAS was validated using eye-tracking data. Data

    used for this validation consisted of four different data sets,

    two of which were collected by academic research labs

    (MIT & York University) and the third and fourth sets were

    collected by 3M. Each data set consisted of eye fixations data

    collected for relatively short periods of time (3-5 seconds) for

    a variety of images (indoor scenes, outdoor scenes, people,

    advertisements, etc.). This provides a measurement as to

    where people will initially look at an image.

    All of the data were collected in a similar way: participants

    were seated in front of a computer monitor and images were

    presented one at a time for a given period of time. In between

    each image there was either a short fixed inter-image pause or

    a wait period in which the subject pressed a button indicating

    that they were ready for the next image. Participants were

    instructed to “freely view “the images during that time (that

    is, they were not given any specific task to complete). During

    the free viewing period, eye-tracking equipment measured

    and recorded where the participants looked at the image.

    Figure 1 provides an illustration of eye-tracking data for a

    single subject. The blue circles indicate the location and

    the size of the circle indicates the dwell time for a particular

    subject. The red “+” in the upper-left illustration of Figure 1

    shows the combined location of all 20 participants who looked

    at this image in the 3M study.

    Validating 3M Data Collection Techniques

    One of the goals of the current study is to show that the

    methods 3M uses to collect data match those of outside

    research institutions. To do this, we collected eye-tracking

    data using the images from a previous study conducted at

    York University in which eye-tracking data from their lab is

    made publically available (http://www-sop.inria.fr/members/

    Neil.Bruce/eyetrackingdata.zip). The York University images

    consisted of 120 images of various indoor and outdoor



  • 3. Response-Operator-Characteristic

    LTo evaluate the predictive performance we calculated ROC

    values using a split-data method in which we calculated how

    well one-half of the subjects fixation data predicted the second

    half of subjects (randomly selected) for each image and then

    compared the predictive power of the 3M VAS predictions

    to that of using human data. To do this we generated two

    predictive maps for each image. One predictive map was

    generated by taking one half of the subjects and generating a

    heat map representation from the subjects’ fixation locations.

    The second heat map was generated by using 3M VAS to

    analyze each image.

    For both sets of predictions (human and 3M VAS) for each

    pixel there is a continuous value associated with the strength

    of the prediction for that pixel. To generate an ROC value

    for each image we varied the threshold of how liberal of a

    prediction would be considered. When the threshold is high,

    the model makes very few predictions as to where attention

    will be allocated, has only a few hits and has very few false

    alarms (see lower left corner of ROC curve in Figure 4).

    However, as the threshold decreases, the area in which the

    model is predicting becomes larger and the model correctly

    predicts more fixations. However, this increased region

    also increases the number of false alarms. To evaluate the

    performance of information available in the Heat Map we

    used the ROC calculation described above.

    ROC takes into account multiple threshold levels and

    measures the number of Hits (correct predictions) and False

    Alarms (incorrect predictions; see Figure 4 for an illustration)

    for each threshold level. After generating the Hits and

    False Alarm rates for multiple thresholds, a single metric

    is generated by calculating the area under the curve. Thus,

    if the model perfectly predicts the data, the prediction will

    have a ROC value of 1.0. Figure 5 (next page) shows the

    distribution of ROC values for the different images for the

    Human-To-Human comparisons for the York University,

    MIT and 3M advertising data. The York University data

    predicted itself with an average ROC value of 0.819. The 3M

    data predicted the York data with an average ROC value of

    0.812. This insignificant difference indicates that the methods

    and procedures used at 3M closely match those used at York

    University. The MIT data predicted itself with an average

    ROC value of 0.89 and the 3M advertising data predicted

    itself with an average ROC value of 0.93. These values

    provide us with the upper-theoretical performance boundary

    by which we will compare 3M VAS performance.

    3M VAS Prediction Efficiency

    The Human-to-Human analysis provides a valuable metric for

    evaluating the efficiency of 3M VAS. Because the Humanto-

    Human analysis provides a way to measure the Theoretical

    Limit of performance — the very best predictive



    Latex – This is the latest and greatest technology in the wrap world and uses less abrasive inks than solvent printers. Latex printers are unique in that they use multiple fans and extremely high-powered heaters to cure the ink as it is being laid down on the vinyl, resulting in prints that are 100% dry immediately after being printed.  Latex printers completely eliminate the need for off-gassing which makes them ideal for the vehicle wrap industry.  Moreover, since latex ink’s composition is naturally more flexible than solvent/eco-solvent ink, it’s ideal for applications requiring vinyl to conform or stretch around curves.


    UV – Ultra Violet (UV) printers are not typically recommended for printing vehicle wraps because the inks’ curing process under UV lights makes the ink brittle when dry.  These printers are great for doing signage and window graphics, but when you take UV printed vinyl and try to stretch and conform it to fit around highly compound curves or bend into deep crevices, there’s a greater risk for ink cracking underneath the laminate.

  • 4. Fixation Biases in Eye-Tracking

    Within the eye-tracking community it is well known that

    when collecting eye-tracking data on computer displays there

    are particular fixation biases that are not determined by the

    content that is on the computer screen. One well known bias

    is known as the Center Bias or the Center Fixation Bias. The

    cause of this bias is due to multiple factors. One known factor

    is that fixating in the center of the screen provides the most

    information about the image (given the resolution fall-off in

    peripheral vision). However, the center bias is due to the fact

    that people are looking at images on a computer screen. 3M

    VAS does not predict how people will look at these images on

    a computer screen but instead how people will look at these

    images in the real world . Using a technique described in

    Zhang et al. (2008) we evaluated both the Human-to-Human

    performance and the 3M VAS performance removing the

    center bias and any other fixation biases that might have

    occurred in the data collection.

    Within the eye-tracking community it is well known that

    when collecting eye-tracking data on computer displays there

    are particular fixation biases that are not determined by the

    content that is on the computer screen. One well known bias

    is known as the Center Bias or the Center Fixation Bias. The

    cause of this bias is due to multiple factors. One known factor

    is that fixating in the center of the screen provides the most

    information about the image (given the resolution fall-off in

    peripheral vision). However, the center bias is due to the fact

    that people are looking at images on a computer screen. 3M

    VAS does not predict how people will look at these images on

    a computer screen but instead how people will look at these

    images in the real world . Using a technique described in

    Zhang et al. (2008) we evaluated both the Human-to-Human

    performance and the 3M VAS performance removing the

    center bias and any other fixation biases that might have

    occurred in the data collection.


  • 5. Modified ROC

    The modified ROC (mROC) calculates performance in much the same way as the standard ROC. As a reminder, the standard ROC calculates the number of hits as a function of the number of false alarms (incorrect predictions) by varying the levels of threshold (threshold independent analysis). The right side illustration in Figure 4 illustrates a sample curve. The mROC, by contrast, evaluates how well the predictions generated for a particular image (e.g., Image-1) predict the

    fixations for that image against the fixations for all of the other images. As with the standard ROC approach, the mROC calculates the percentage of hits for the target image (i.e., correctly predicting the fixations for the target image for a particular threshold) against the percentage of hits for the

    fixations for all other images in the study. It is well known that when collecting data on a computer

    there is a bias for fixating in the center of the screen (Tseng et al. 2009). There are a number of reasons for this center bias, which were discussed and evaluated by Tseng et al. (2009), that we will not go into in this manuscript. Removing this center bias from the analysis is important since the performance of the model will be reduced because the model predicts what people will see in a real environment (free viewing with no frame), not how perform in an eyetracking

    study on a computer. It should be noted that one way to improve the model’s performance is to put an explicit center bias within the model’s predictions. However, this improves the model’s performance for predicting fixations on a computer monitor but not necessarily in the real world.

    One way to think about the mROC calculation is that the mROC evaluates how well the model is able to predict the unique fixations for a target image relative to all other images. More accurately, the more unique the correctly predicted fixation positions for a particular image, the more

    weight that is given to the score. Therefore, if there are numerous fixations in the center of the computer screen and the model correctly predicts these fixations, these scores will

    be weighted less than a prediction made to fixations where, overall, there were very few fixations across all of the images.



A study from 3M shows that a business would have to spend over $130,000 on television ads to generate the annual impressions of a $3,500 wrap!


One of the most cost effective forms of advertising available.  In a city of 50,000, a wrap can get over 8 Million views per year!  Works for you  24/7  One time advertising  expense, unlike other traditional media.  Can accommodate most budgets with full and partial wraps.






Wraps come in all different shapes & sizes. Wraps start as large sheets of adhesive backed vinyl film designed specifically for vehicle application made available in a variety of styles & finishes. Some wraps are a full color change in matte, satin or gloss finishes, as well as chrome, “carbon fiber” look, and beyond. Vehicle wraps can also consist of printed designs containing imagery, photos, graphics, or branded elements to represent a business or group, or as “accents” on vehicle features like trim, spoilers, hoods, roofs, and trunks. Truly, the possibilities are endless; recent advancements in vinyl production technology has also opened up the possibilities to textured vehicle wraps like leather, alligator, snake skin, or holographic & 3D films!




Every vehicle & application is different, so price really depends on the scope of your vehicle wrap project. A complete wrap includes door jambs, inner door frames, and sills — this level of vehicle coverage is recommended for those who desire a full color change, and is obviously the most costly. Partial wraps without jambs, frames, and sills are cheaper, but ultimately will not have the same illusion as full coverage applications. All of our prices listed are base prices, and will vary depending on the vehicle type, surface complexity, and film used. For a more precise quote, contact us with more information about your wrap project.

Click here for a general Price!

NOTE: prices do not include full vehicle disassembling or re-assembling, or removal of previous films, wraps, and clear bra.




No, but improper installation or removal can! A bad install means your vehicle’s paint might get damaged from razor blade scratches, and poor adhesion that ultimately leads to wrap failure. Also, if the wrong type of film is used during the application like a sign vinyl with extra strength adhesive or low flexibility, it can strip your paint and cause serious damage if & when it is not removed with care!




Wrap life span really depends on how well the film is cared for, how much day to day stress the wrap gets from road driving, and what quality of film was used to do the installation. Some films last longer than others, and the thicker they are the more damage they can take. We’ve wrapped vehicles that see plenty of road & track time, and their films have survived more that 5 years without needing parts re-installed or failing entirely.




Most definitely. While it is a bit of a tedious process to undertake, we offer wrap removal services as well as installation, and would be happy to assist. Contact us with more information about what type of wrap you would like to have removed & we’ll help you get back to fresh paint!

Can you install my vinyl for me?

Typically we don’t install any film we’re not already stocking. There are a lot of knock off films out there that aren’t designed for vehicle wrap applications, so if you want help making a selection just contact us for a recommendation!




No. With primed and unpainted vehicles or body kits, the surface texture does not allow for proper adhesion of air egress vinyls, and makes the wrap prone to failure & lifting.




Proper wrap care starts with the basics. Keeping your vehicle’s surface clean is the primary concern, so frequent hand washing to eliminate surface contaminates is necessary if you want to keep your wrap from getting stained or damaged from road grime. For those of you with a gloss free exterior, we offer wrap waxing services for matte & satin films to help keep them protected as well as maintain a nice soft, flat finish.




Primarily, the film thickness. Matte opaque options have become more and more popular for those wanting a complete color change, while our satin clear applications are more tailored to those who simply want to change their factory paint finish while also adding a layer of protection.




Most opaque and textured wrap films are thin and pliable so they can stretch easily during application. Unfortunately, thinner films are more prone to failure, and typically don’t offer much paint protection. However, some wrap films are so thick that they can add a layer of protection – specifically, our satin clear applications. Typical matte colored film is around 2-4 MILS thick and wont bare too much of an impact from large rocks or chunky road debris, but satin clear is on par with standard clear bra protection at around 7-9 MILS thick, keeping your paint that much safer.




We typically don’t stock much vinyl other than what is necessary for current projects. Since we work with such a diverse range of clientele with a wide array of tastes and styles, we install just about every type of film on the market (and some custom ones too!) 3M, Avery, Oracal, Hexus, KPMF, you name it, we’ve used it. If there’s a specific film you want to use, let us know and we’ll find it for your vehicle at the best price available. We can even special order a one-off film, create a custom printed design, or track down something unique if you really want to break away from the pack!




As the age old saying goes, you get what you pay for. Our prices reflect the quality of work that goes into each installation, and our installs are incredibly thorough & time consuming. Speedy & cheap installs typically lead to cars being bladed up, films not being properly applied or given enough time to breathe, and ultimately a shorter life span for the wrap. Also, most installers will not offer any sort of warranty if the wrap film fails or was applied improperly. Not only do we make sure that your vehicle’s exterior is spotlessly clean to ensure proper adhesion and no film failure, but we also go the distance in disassembling your vehicle prior to wrapping. That means no unnecessary seams, no bad cuts, or messy edges. If the film we supplied fails, bring it in and we’ll make it right until you’re satisfied with the results.




If the wrap is our primary focus, turn around time can be as short as a week or less depending on complexity. Stripes, graphics & accent pieces can typically be completed in 24 – 72 hours, but exterior wraps or full color changes with door jambs will take considerably longer. Since we’re a fully functioning detail shop with more than just wraps on our plate, installation times will vary depending on the project and when it’s scheduled. If you’re on a deadline or want to expedite your install, just let us know what the time frame is and we’ll factor it in when we set up your appointment.




There are pros & cons to both wrapping or painting your vehicle, and normally a high quality paint job will cost just as much if not more than a high quality wrap. A cheap paint job is rarely a good thing, and preserving original paint is something we value, so we tend to side with wrapping. While it might be “cheaper” to do an easy respray on your vehicle for a change of color, it will ultimately damage the value and make it harder to resell down the line. A wrap can be removed easily, doesn’t run the risk of inconsistencies that come with a cheap paint job, and can potentially increase the resale value to prospective buyers if they plan on keeping it wrapped.




We don’t regularly, but discounting vehicle wraps is something we take into consideration for appropriate parties who would like it known where they had their wrap installed. If you are interested in representing and promoting the NorthWest Auto Salon brand as part of your color change, printed wrap, or custom graphics in a new or innovative way, please send us a complete proposal with details about the vehicle, wrap design, film choices, project affiliates, and a schedule of where the vehicle will be on display or in competition.




In most cases, yes. While the jury is still out on the legality of vehicle wraps in most states due to their recent rise in popularity, we have heard of action being taken to limit certain films from road use by regulating their production size. Varieties of chrome & reflective films are intended for off road use only or for minor graphical additions as opposed to total car coverage, but there isn’t a current state law barring you from changing the color of your car via vehicle wrap that we know of, or requiring you to re-register your vehicle in a new color.




For starters, we were one of the first detail shops in the greater Pacific North West let alone the United States to performed a full vehicle wrap using matte black film. We’ve had more experience with total coverage vehicle wrapping than any other detail shop in the business, and have worked on wrap projects spanning a wide array of cars & clientele from the most average to the most exotic. Our in house design team produces cutting edge aesthetics that set the trends for the rest of the wrap industry, and our access to exclusive materials & manufacturers is unmatched. Having your vehicle wrapped by

VIP WRAPS.PRO ensures a level of quality & support that simply cannot be found anywhere else!




Most definitely. Dirt and contaminants on your exterior would be trapped under the film, posing greater harm to your paint and eventually causing the wrap to fail if they were not cleaned off before installation. Thus, we must require a full 24- step hand wash and clay bar decontamination before the wrap is applied.




There are a few courses of action we can take depending on how your wrap gets damaged, where the damage occurred, and how extensive it is. Worst case scenario, the wrap must be removed entirely, but more than likely we will be able to pull a section of the wrap and replace it with a new piece of film that was left over from your installation. Depending on the design or layout of your wrap and where the damage is located, we may also need to reprint graphics or simply patch a spot in.




Printed, branded, and customized vehicle wraps are something we specialize in. If you already have a design mocked up and simply need to get film printed and applied, or you want us to take it all the way from concept to completion, we would be glad to assist in bringing your personal or company fleet vehicles up to spec.




Thank you for choosing VIPWRAPS.PRO, the bay area's number one Vinyl wrap company. To request an estimate or discuss your project with our Vinyl Wrap Certified professional, call or email us

Submitting Form...

The server encountered an error.

Form received.


 © 1999, all rights reserved.  Designed & Marketed by VIP RE-STYLERS Services             | Privacy Policy | Home | About | Services | Contact