CURR TPCS IN COGNITIV NEUROSCI
CURR TPCS IN COGNITIV NEUROSCI PSY 394U
Popular in Course
verified elite notetaker
FLGR 1010 - 005
verified elite notetaker
verified elite notetaker
verified elite notetaker
verified elite notetaker
verified elite notetaker
Popular in Psychlogy
This 33 page Class Notes was uploaded by Marco Wolf on Monday September 7, 2015. The Class Notes belongs to PSY 394U at University of Texas at Austin taught by Wilson Geisler in Fall. Since its upload, it has received 7 views. For similar materials see /class/181797/psy-394u-university-of-texas-at-austin in Psychlogy at University of Texas at Austin.
Reviews for CURR TPCS IN COGNITIV NEUROSCI
Report this Material
What is Karma?
Karma is the currency of StudySoup.
You can buy or earn more Karma at anytime and redeem it for class notes, study guides, flashcards, and more!
Date Created: 09/07/15
V39suzl Neurnscience Hnw many nixEIS make an image scholarour Muunsuim cmm Page 1 of 32 Visual Neuroscience Title How many pixels make an image Antonio Torralba Computer Science and Arti cial Intelligence Laboratory Massachusetts Institute of Technology Telephone 617 324 0900 Fax 6172586287 Email Torralbacsailmitedu Address for correspondence 32D432 32 Vassar Street Cambridge MA 02139 Short title How many pixels make an image Number of manuscript pages 14 number of tables 0 number of gures 9 Visual Neuroscience Page 2 of 32 Abstract The human Visual system is remarkably tolerant to degradations in image resolution human performance at natural image categorization remains high whether low resolution images or multimega pixel images are used This raises the question of how many pixels are required to form a meaningful representation of an image and identify the objects it contains In this paper we show that very small thumbnails images at a spatial resolution of 32x32 color pixels provide enough information for identifying the semantic category of real world scenes Most strikingly this low resolution permits to report with 80 accuracy 45 of the objects that the scene contains despite that some of these objects are unrecognizable in isolation The robustness of a short code for describing semantic content of natural images could be an important asset to explain the speed and efficiently at which the human brain comprehends the gist of Visual scenes Keywords scene recognition object segmentation gist thumbnails natural images blobs Page 3 of 32 Visual Neuroscience Intro du ction In the images shown in Figure 1a we can easily categorize each picture into scene classes a street an of ce etc We can also recognize and segment many ofthe objects in each image Interestingly though these pictures have only 32 X 32 pixels the entire image is just a vector of 3072 dimensions with 8 bits per dimension yet at this resolution the images seem to already contain most of the relevant information needed to support reliable recognition of many objects regions and scenes This observation raises the question of how many pixels are needed to form a meaningful image In other words what is the minimal image resolution at which the human visual system can reliably extract the gist of a scene the scene category and some of the objects that compose the scene The gist ofthe scene Friedman 1979 Oliva 2005 Wolfe 1998 refers to a summary of a semantic description of the scene ie its category layout and a few objects that compose the scene Such a summary may be extracted from very lowresolution information Oliva amp Schyns 2000 Oliva amp Torralba 2001 and therefore can be computed very efficiently Low dimensional image representations and short codes for describing images can be important to explain how the brain can recognize scenes and objects very fast VanRullen amp Thorpe 2001 have suggested that given how fast recognition happens 150ms after stimulus onset recognition might be driven at rst by feedforward mechanisms in which neurons only have time to re one or two spikes They discuss that even with such a small amount of information and when only a small Visual Neuroscience Page 4 of 32 fraction of the neurons re one spike it is possible to perform challenging recognition tasks such as detecting the presence of animals in natural images Bar 2007 suggests that lowspatial frequencies activate expectations that will facilitate bottomup processing In Torralba et a1 2007 a low dimensional image representation is used to guide attention incorporating information about the scene context and task constraints The problem studied in this paper is to nd what is the minimal resolution required to perform scene recognition and object segmentation in natural images Note that this problem is distinct from studies investigating scene recognition using very short presentation times and perception at a glance Greene amp Oliva in press Joubert el al 2005 Schyns amp Oliva 1994 Oliva amp Schyns 1997 Potter et al 2002 Intraub 1981 Rousselet et al 2005 Thorpe et al 1997 FeiFei et al 2007 Rousselet et al 2005 Here we are interested in characterizing the amount of information available in the image as a function of the image resolution there is no constraint on presentation time Studies on face perception Bachmann 1991 Harmon amp Julesz 1973 Schyns amp Oliva 1997 have shown that when a picture of a face is downsampled to a resolution as low as 16x16 pixels humans are able to perform various face recognition tasks reliably ie identity gender emotion Remarkable performance with low resolution pictures is found also on scene recognition tasks Oliva amp Schyns 2000 Castelhano amp Henderson 2008 In this work we explore what is the minimal image resolution required to perform scene recognition and object segmentation We will show that at very low resolutions a large number of tasks including hard ones such as object segmentation can be reliably performed Page 5 of 32 Visual Neuroscience Patches objects and scenes Scene pictures are quite different to patches randomly extracted from images Fig 1 Figure 1b shows 32x32 pixel patches randomly selected from natural images A number of studies Olshausen amp Field 1996 Lee et al 2003 Chandler amp Field 2006 have been devoted to characterize the space of natural images by studying the statistics of small image patches such as the ones shown in Fig 2b Those studies helped to understand the receptive elds of visual neurons in early visual areas Olshausen amp Field 1996 However many of these patches do not contain enough information to be recognized as part of a speci c object or region as they contain at surfaces or insufficient edges Figure 1c show tight crops of objects rescaled at 32x32 pixels These are the kind of images used in computer vision to train object detection algorithms Olshausen et al 1993 proposed an attentional system that selected 32x32 windows around regions of interest and argued that this was enough for recognizing most objects Tight object and crops of objects without background have also been the focus of many studies in visual cognition Those studies have focused mainly in the study of faces using image resolution as a way of controlling the amount of global and local information available Figure 1a corresponds to full scenes what a human standing on the ground and looking at a wide scene would typically see scaled to 32x32 pixels These scenes contain many objects which amazingly are still recognizable despite the fact they occupy just a few Visual Neuroscience Page 6 of 32 pixels Scene pictures include biases introduced by the way that humans tend to take pictures Although this could be considered as a bias in the dataset we think that such a bias is due to observer constraints and it should be taken into account when coding images Materials and methods The images used for this study were drawn from the scenes dataset from Oliva amp Torralba 2001 and the LabelMe database Russell et al 2008 In order to cover a large variety of different scenes we collected 240 images distributed across 12 scene categories as follows 6 outdoor categories street highway seaport forest beach and mountainous landscape and 6 indoor categories corridor kitchen bathroom living room bedroom and office All the images were originally of size 256x256 pixels For each image we generated lowresolution images at 4x4 8x8 16x16 32x32 64x64 and 128x128 pixels In order to reduce the resolution of each image we first apply a low pass binomial filter to each color channel with kernel 1 4 6 4 1 and then we downsample the filtered image by a factor of 2 Next each pixel of the lowresolution images is quantized to 8 bits for each color channel For visualization the low resolution images are upsampled to 256x256 pixels Previous studies use a Gaussian filter in order to blur the images The problem with using a Gaussian filter without downsampling the image is that it is difficult to evaluate the Page 7 of 32 Visual Neuroscience exact amount of information that is available to the observer By f1rst subsampling the image it allows to establish a clear bound on the amount of visual information available In this paper we will use the size of the downsampled image as a measure of the amount of visual information that is available on the blurred images Scene recognition Experiment There were 28 na39139ve observers age ranging from 18 to 40 years old that took part in the scene recognition experiment They all gave informed consent The experiment had two conditions color images and grayscale images 14 observers participated in the color condition and 14 in the grayscale condition Each image was shown at one of the 6 possible resolutions 4x4 8x8 16x16 32x32 64x64 and 128x128 pixels All images were upsampled to 256x256 pixels for display and shown only once to each observer The procedure consisted in a 12 alternative choice task each image was categorized as belonging to one ofthe 12 possible scene categories The image was displayed on the screen until the participant made a choice Each participant saw a total of 240 images in a random order Results Visual Neuroscience Page 8 of 32 Figure 3 provides the overall pattern of results on the scene categorization task for color and grayscale images as a function of image resolution Below the graph the top row of images illustrates the number of pixels at each resolution The lower row shows the images that were shown to the participants during the experiment When images were shown at 128x128 pixels performances were at ceiling at 96 correct recognition rate A few of the scene images are ambiguous in terms of a unique scene category like a road with a mountain which could be classified as a mountainous landscape or as a highway therefore 100 recognition rate is impossible at this task Chance level in this task is at 83 At a resolution of4 x 4 pixels performance for grayscale images was 9 and was not significantly different from chance tl3ltl but color was signi cantly higher than grayscale performance t2638 plt0001 with a correct rate of 184 Scene recognition performance remains high even at relatively low spatial resolutions Oliva amp Schyns 2000 Participants had an 818 correct recognition rate when color images were presented at a resolution of 32x32 pixels Performance over 80 is found here for resolution slightly higher than the results found in Oliva amp Schyns 2000 who used a Gaussian filter and reported the frequency of 50 cutoff instead of downsampling the image For grayscale images performance was at 73 at 32x32 pixels As we lower the resolution performances drop There is a significant improvement in recognition performance when color is present Oliva amp Schyns 2000 Castelhano amp Henderson 2008 Rousselet et al 2005 for low resolution images Page 9 of 32 Visual Neuroscience The results presented in gure 2 were averaged over all scene types However it is important to note that different types of images will lose information at different rates as their resolution decreases The difference between the resolution requirements of different images is illustrated in Figure 3 In this gure images are sorted by image resolution needed for recognition As the figure illustrates some images can be recognized at extremely low resolutions even when only 8x8 pixels are available as the images on the left in Figure 3 while others require higher resolution Figure 4 shows the recognition performances averaged over three groups of scenes indoor outdoor manmade and outdoor natural scenes Each of these superordinate scene groups has different requirements in terms of resolution and contribution of color information for the scene categorization task Indoors are the hardest classes of scenes to be recognized First we note that the contribution of color information for indoor scenes is marginal compared to the other scene types Recognition rate at 32x32 pixels is at 778 correct for color images and at 678 for grayscale images On the other hand outdoor manmade scenes have a 922 correct recognition rate when presented in color at 32x32 pixels and 788 when presented in grayscale For outdoor scenes the inclusion 1 of color information provides a great imp v t on the quot39 rate to r indoor scenes A similar trend is observed for outdoor natural scenes although for natural scenes recognition rate remains high even for very low resolutions at 8x8 pixels performance is still around 651 Visual Neuroscience Figure 5a provides the recognition performance as a function of image resolution for the 12 scene categories used in the experiment Also note that adding color information does not improve recognition performance with respect to grayscale images for most indoor categories On the other hand color provides a dramatic increase in recognition performance for natural environments especially at very low resolutions The contribution of color information is made clear in figure 5 Figure 5 shows the average of all the images that belong to each of the 12 scene categories The top row represents the indoor categories Only the corridor category can be easily recognized in this average image The bottom row represents the 6 outdoor scene categories used in the experiments Outdoor categories have a wider distribution of colors that are good predictors of the scene category In a series of experiments Oliva amp Schyns 2000 showed that diagnostic colors are important in many outdoor categories particularly natural landscapes as color of surfaces is a diagnostic feature of the scene category see also Goffaux et al 2005 Rousselet et al 2005 However color becomes less important to differentiate among manmade environments were the color of objects and surfaces are often accidental An important observation is that not all the images can be correctly interpreted at very low resolutions In order to study the distribution of resolutions needed to reach 80 recognition rate on single images we performed an item analysis As we do not have enough data to estimate the performance at each resolution independently we use a psychometric function to estimate the relationship between recognition performance and image resolution For each image we use logistic regression to fit the psychometric Page 10 of 32 Page 11 of 32 Visual Neuroscience function Klein 2001 relating probability of correct recognition as a function of image resolution we use this function as it seems to fit the observed data However there are several other choices that will also be valid This function is defined by only two parameters ab therefore it can be estimated with few samples Pcorrect resolution a b l 1expa b39 logresolution The parameters of this psychometric function are estimated using maximum likelihood Once the parameters are estimated we nd for each image what is the minimal resolution needed to reach 80 recognition rate Figure 6 shows the histogram of image resolutions that reach 80 recognition rate with color and grayscale images Only a very small percentage 3 with color ofthe images need high resolution images 128x128 pixels For color images more than 80 ofthe images need 32x32 pixels or less to be recognizable Object segmentation Experiment 15 participants perform a total of 1195 trials For each trial an image was randomly selected from the set of 240 images described before Each image was presented at one of Visual Neuroscience Page 12 of 32 the 6 possible resolutions 4x4 8x8 16x16 32x32 64x64 and 128x128 For this experiment images were presented in color Participants were asked rst to provide the scene category of the image this part of the experiment was identical to the scene recognition experiment described before then participants were asked to segment using a drawing tool as many objects and regions as they could recognize on each image They traced and segmented one regionobj ect at a time entered its possible name and then traced a second object named it etc The 15 participants annotated atotal of 5706 objectsregions Once the experiment was concluded we run a validation procedure to decide which objects were correctly recognized For this ground truth validation each object was shown at the original resolution together with the annotation provided by one of the participants For the validation the information about the level of blur shown to the participant was not shown to avoid any bias The images and participants were randomized for the validation stage The validation was performed by the author Results Figure 7a shows an example image and the segmentations produced by 6 participants at different resolutions As the resolution of the image increases the participants reported more objects and the reported objects had a higher probability of being correct The question that we address here is to nd what is the minimal image resolution needed so Page 13 of 32 Visual Neuroscience that participants can extract the information equivalent to the gist of the scene Oliva 2005 and Wolfe 1998 argue that the gist ofthe scene might be composed of a coarse representation of the scene layout and a list of 4 or 5 objects Figure 7bc summarizes the object recognition results for each of the three super ordinate scene categories Figure 7b gives the number of reported objects as a function of image resolution As more resolution is available more objects become recognizable The number of objects reported seems to grow logarithmically with the image resolution for the three superordinate scene categories Participants reported fewer objects for natural environments than for manmade environments Figure 7c gives the recognition rate for the reported objects At a resolution of 32x32 pixels with color images participants reported on average 5 across all the superordinate scene categories objects with 80 correct recognition rate Figure 7d shows the distribution of sizes for the reported objects For all image resolutions most of the reported objects had an area that covered between 18 and 12 of the image area Figure 7e shows that the percentage of correctly reported objects did not vary a lot between scales despite the large variation on object sizes Figure 8a shows several examples of images at 32x32 pixels and the segmentations provided by the participants Figure 8b shows some of the reported objects isolated from the scene Some of the objects are defined by just a few blobs and recognition is only reliable when they are immersed in the scene Visual Neuroscience Page 14 of 32 At very low resolutions recognition is heavily driven by contextual relationships between objects This point is quanti ed in gure 9 Figure 9 shows that there is a signi cant interaction between performances on the scene recognition task and the object segmentation task Figure 9 splits the object recognition results depending on whether participants identi ed correctly the scene prior to the object segmentation task The number of objects reported did not change when participants missclassi ed the scene category On average for 32x32 images participants reported 5 objects independently on whether they assigned the correct or incorrect scene category to the scene However object recognition performance drops dramatically for low resolutions At a resolution of 32x32 pixels the reported objects were correctly recognized 89 of the times when the scene was correctly identi ed Performances dropped to 53 when the reported scene was wrong Failing to recognize the scene context has a major effect on object recognition performances at low image resolutions At highresolution the effect is not so important The same trends are observed when the performances are analyzed for the three superordinate scene groups indoors manmade outdoor and natural outdoor Conclusion A lot of research effort has been devoted to understand how early visual areas in the brain process ne image structures such as edges junctions textures shape descriptors etc However very low resolution blobs particularly colored blobs provide an incredible amount of information that could guide high the processing of highresolution image detail Schyns amp Oliva 1994 Bar 2007 In this paper we have explored two tasks Page 15 of 32 Visual Neuroscience scene and object recognition as a function of image resolution We have shown that 32x32 color images are already well formed pictures with meaningful and recognizable global and local structures Strikingly for more than 80 of the images used in our experiments the gist of the scene the scene category and the identity and localization of 45 objects that compose the scene is available with an image resolution of just 32x32 color pixels Even the lowest spatial frequency channels provide enough information for reliable recognition of the scene category which can in turn be used to facilitate subsequent analysis of local regions and smaller objects in the scene How is recognition even possible at such low resolutions At a resolution of 32x32 pixels most regular image features related to textures junctions contours are not available or are very weak Therefore blob shapes and their spatial relationships become central for understanding the image content A coding of the contextual relationships between regions is mandatory in order to achieve good recognition performance The robustness of a short code for describing semantic content of natural images could be an important asset to explain the speed and efficiently at which the human brain comprehends the gist of visual scenes Acknowledgments Funding for this research was provided by NSF Career award HS 0747120 Visual Neuroscience Page 16 of 32 References Bachmann T 1991 Identi cation of spatially quantized tachistoscopic images of faces How many pixels does it take to carry identity European Journal of Cognitive Psychology 3 857103 Bar M 2004 Visual objects in context Nature Neuroscience Reviews 5 617629 Bar M 2007 The Proactive Brain Using analogies and associations to generate predictions Trends in Cognitive Sciences 117 280289 Castelhano MS amp Henderson J M 2008 The In uence of Color on Perception of Scene Gist Journal of Experimental Psychology Human Perception anal Performance 34 660675 Chandler D M amp Field D J 2006 Estimates of the information content and dimensionality of natural scenes from proximity distributions JOSA 249227941 Fei Fei L Iyer A Koch C amp Perona P 2007 What do we perceive in a glance ofa realworld scene Journal ofVision 71 129 Page 17 of 32 Visual Neuroscience Friedman A 1979 Framing pictures the role of knowledge in automatized encoding and memory of gist Journal of Experimental Psychology General 108 316355 Goffaux V Jacques C MourauX A Oliva A Rossion B amp Schyns PG 2005 Diagnostic colors contribute to early stages of scene categorization behavioral and neurophysiological eVidences Visual Cognition 12 878892 Greene MR amp Oliva A in press Recognition of natural scenes from global properties seeing the forest without representing the trees Cognitive Psychology Harmon L D amp Julesz B 1973 Masking in Visual recognition Effects oftwo dimensional filtered noise Science 180 11941197 Intraub H 1981 Rapid conceptual identification of sequentially presented pictures Journal of Experimental Psychology Human Perception and Performance 7 604610 Joubert 0 Rousselet G Fize D amp FabreThorpe M 2007 Processing scene context fast categorization and object interference Vision Research 47 32863297 Klein S A 2001 Measuring estimating and understanding the psychometric function A commentary Perception amp Psychophysics 63 8 14211455 Lee A B Pedersen K S amp Mumford D 2003The nonlinear statistics of high contrast patches in natural images Int J Comput Vision 5413837103 Visual Neuroscience Page 18 of 32 Navon D 1977 Forest before the trees the precedence of global features in visual perception Cognitive Psychology 9 353383 Oliva A amp Schyns PG 2000 Diagnostic colors mediate scene recognition Cognitive Psychology 41 1767210 Oliva A amp Torralba A 2001 Modeling the shape ofthe scene a holistic representation of the spatial envelope International Journal of Computer Vision Vol 423 145 175 Oliva A 2005 Gist of the scene In the Encyclopedia of Neurobiology of Attention L Itti G Rees and JK Tsotsos Eds Elsevier San Diego CA pages 251256 Oliva A amp Torralba A 2007 The role of context in object recognition Trends in Cognitive Sciences vol 1112 pp 520527 Olshausen B A et a1 Anderson C H and Van Essen D C 1993 A neurobiological model of visual attention and invariant pattern recognition based on dynamic routing of information J Neuroscience 13 47004719 Olshausen BA Field DJ 1996 Emergence of SimpleCell Receptive Field Properties by Learning a Sparse Code for Natural Images Nature 381 607609 Page 19 of 32 Visual Neuroscience Potter MC 1975 Meaning in Visual scenes Science 187 965966 Rousselet G A Joubert O R FabreThorpe M 2005 How long to get to the quotgistquot of realworld natural scenes Visual Cognition 126 852877 Russell B Torralba A Murphy K amp Freeman W T 2008 LabelMe a database and webbased tool for image annotation International Journal of Computer Vision 773 157 173 Sinha P Balas BJ Ostrovsky Y amp Russell R 2006 Face recognition by humans 19 results all computer Vision researchers should know about Proceedings of the IEEE Vol 94 No 11 19481962 Schyns PG amp Oliva A 1994 From blobs to boundary edges Evidence for time and 1 1 u n 11 spatial a1 scene 39 1 Science 5 195200 Thorpe S Fize D amp Marlot C 1996 Speed of processing in the human Visual system Nature 381 520522 Torralba A Oliva A Castelhano M amp Henderson J M Contextual Guidance of Attention in Natural scenes The role of Global features on object search Psychological Review 2006 1134 p 766786 Visual Neuroscience Page 20 of 32 Torralba A Fergus R amp Freeman W T 2008 80 million tiny images a large dataset for nonparametric object and scene recognition IEEE Transactions on Pattern Analysis and Machine Intelligence VanRullen R amp Thorpe 8 2001 Rate coding versus temporal order coding what the retinal ganglion cells tell the Visual cortex Neural Computation 13 125583 Van Rullen R amp Thorpe 8 2001 The time course of Visual processing from early perception to decision making Journal of Cognitive Neuroscience 134 454461 Wolfe J M 1998 Visual Memory What do you know about what you saw Current Biology 8 R303R304 Page 21 of 32 Visual Neuroscience Figure 1 Scenes patches and objects all at 32x32 pixels Note how rich the scenes and objects are in comparison with the image patches Figure 2 Scene recognition as a function of image resolution Error bars represent one standard error of the mean obtained from 12 participants for each condition The vertical axis represents the correct recognition rate and the horizontal axis corresponds to the image resolution in a logarithmic scale The black horizontal line represents chance level The two rows of images illustrate the amount of information available at each resolution The top row shows the downsampled images at each resolution from 4x4 to 128x128 pixels and the second row shows the images upsampled to 256x256 pixels that were shown to the participants Figure 3 Images sorted by the amount of resolution required for becoming recognizable Each row shows images that had been downsampled to a resolution of 12x12 18x18 28x28 42x42 and 64x64 pixels The two images on the left a corridor and a beach are correctly categorized by most of the participants even at the lowest resolutions The two images on the left require very high resolution in order to become recognizable an office and a bedroom The two images on the center need around 32x32 pixels in order to be recognized by most of the participants Easy images are formed by few surfaces and had diagnostic spatial layouts Visual Neuroscience Page 22 of 32 Figure 4 performances for the 12way classi cation task averaged over indoor scenes 6 categories outdoor man made environments street sea port and highway and outdoor natural landscapes beach forest mountainous landscape Error bars correspond to the standard error Figure 5 A Performance on scene recognition for each scene category chance is at 83 correct recognition rate with grayscale and color images Error bars correspond to the standard error B Average images of all the scene pictures that belong to each scene category Each average is obtained and the pixel wise average of the 20 images in each group The categories shown are from left to right and top to bottom bathroom bedroom corridor kitchen living room office seaport highway street beach forest and mountainous landscape Figure 6 Distribution of image resolutions needed to reach 80 scene recognition rate for color and grayscale images When color information is available there are 20 of images that even at a resolution of 8X8 pixels already reach 80 recognition rate of the scene categorization task while only 11 of the images can be recognized at that resolution when shown in gray scale Page 23 of 32 Visual Neuroscience Figure 7 as we increase the resolution participants report an increasing number of objects The number of objects reported seems to increase logarithmically with respect to the image resolution B From those objects the object correctly recognized also increases with resolution reaching 80 around 32x32 resolution images C D Average number of reported objects per image as a function of object size measured as the proportion of the image occupied by the object For all image resolutions most of the reported objects cover an image area in the interval 1 8 1 4 of the total image size This is between 12 and 25 of the image size E Recognition rate for reported objects as a function of their size on the image Figure 8 Images at a resolution of 32x32 pixels and the segmentations provided by the participants Figure B shows some of the recognized objects cropped Many of those objects become unrecognizable once they are extracted from the surrounding context Figure 9 Performance on object recognition as a function of whether the scene was correctly identified or not A Number of objects reported as a function of image resolution B Recognition rate for reported objects as a function of image resolution Visual Neuroscience Page 24 of 32 l7 c gelmes small thumbnails 106x123mm 600 x 600 DPI Page 25 of 32 percent correct Visual Neuroscience 1396 3392 Resolution pixels 168x124mm 600 x 600 DPI Visual Neuroscience Page 26 of 32 175x162mm 600 x 600 DPI Page 27 of 32 Recognition rate indoor Visual Neuroscience outdoor man made outdoor nature 39 grayscale color 6O Recognition rate 20 a 0 Recognition rate 5 20 4 8 16 32 64 128 Resolution pixels 4 8 16 32 64 128 Resolution pixels 196x68mm 600 x 600 DPI 4 8 16 32 64 128 Resolution pixels Visual Neuroscience Page 28 of 32 bathroom bedroom corridor kitchen office 80 50 recognition rate 10 481632 128 481632 128 forest mountain 481632 128 481632 128 4 816 32 128 4 816 32 128 4 816 32 128A 4 816 32 128 198x145mm 600 x 600 DPI Page 29 of 32 Percentage of images that reach 80 recognition rate 50 40 30 20 10 Visuai Neuroscience 16 Image resolution pixels 32 64 128 169x139mm 600 x 600 DPI Visual Neuroscience 4 r 8 16 32 64 Bedroom A 0 Average number of reported objects 53 5339 0 f f 4gt ax on d N 4gt Ox Average number of reported objects 0 N Beach Bedroom Bedroom 1 n quot39 gt 1 Bedroom Bedroom window lamp 1 night tablet 7 bed ndoor Outdoor manmade quotquot Outdoor nature m100 lt3 9 39Q t g 80 a r 1 o D e e 60 r o I zk i g 0 7 IA g r E 40 xi 0 r g a Indeor g 20 W Outdoor manmade Outdoor nature 5 D 0 16 32 64 128 4 Resolution B 100 8 16 32 64 128 Resolution C 80 o o o Percentage of correctly reported objects l 3 S 0 2 le Relative object area 186x266mm 600 x 600 DPI 4 LI 1 32 16 16 8 8 Relative object 1 4 m rea Page 30 of 32 Page 31 of 32 32x32 Visual Neuroscience sink Corner oftub mirror r curtain sink corner of tub cabinet responseseapon big 7095 boat boat boat sailboat boat boat t 7 boat boat boat Tb water bndge building responsestreet tree sky building 6 Hbrldge tree ighay 39car car highway 210x109mm 600 x 600 DPI Average number of reported objects Visual Neuroecience a Correct scene category Wrong Scene category 4 8 1396 3392 64 128 Resolution 100 60 40 Percentage of correctly reported objects 0 80 20 r t l I I I I I I Correctiscene category Wrong fscene category 4 184x93mm 600 x 600 DPI 8 1396 z 64 128 Resolution Page 32 of 32
Are you sure you want to buy this material for
You're already Subscribed!
Looks like you've already subscribed to StudySoup, you won't need to purchase another subscription to get this material. To access this material simply click 'View Full Document'