33 research outputs found
Poster presentation: The time course of image memorability
No full textSome images we see stick in mind, while others fade. Recent studies of visual memory have found remarkable levels of consistency for this inter-item variability across observers. This suggests that memorability can be considered an intrinsic image property. Most of these studies quantified the memorability of an image as the proportion of participants recognizing it in a repeat-detection memory task. The retention interval in this task is typically quite short (e.g., 4 min.), although one
previous study increased it to 40 min. and found some evidence for the consistency of image memorability over time. The current study sought to further evaluate this consistency with a more traditional visual long-term memory task. Participants studied 342 previously quantified images and completed a first recognition test immediately after. Two additional recognition tests followed one day and one week later. A Latin square design enabled us to quantify the memorability of each image at each retention interval without testing participants for the same image twice. Our
memorability scores show levels of consistency across observers in line with those reported in previous research. They correlate strongly with previous quantifications (rhos between .52–.76 at the shortest interval) and appear stable over time (rhos between .52–.69).sponsorship: PhD Fellowship of the Research Foundation - Flanders (FWO), granted to Lore Goetschalckx (Grant 1108116N).
Methusalem grant awarded to Johan Wagemans by the Flemish Government (METH/14/02)status: Publishe
Poster presentation: Towards a better understanding of image memorability
No full textSome images stick in mind, while others fade. Recent studies have found remarkable levels of consistency for this inter-item variability across observers, suggesting memorability can be considered an intrinsic image property. Besides evident practical implications, a better understanding of this property might elucidate the interaction between perception and memory. My poster presents the results of two initial studies of a broader project. The first sought to further evaluate the consistency with an old/new-recognition memory task consisting of a study phase and a delayed test phase, instead of the repeat-detection task used up till now. We tested three retention intervals (20 min, 1 day, 1 week) and found high levels of consistency across observers at all three. Moreover, our memorability rank scores correlated well with those previously reported and stayed relatively stable across retention intervals. The second investigated the relation between memorability and inter-item variability in a rapid-scene categorization task. Participants briefly saw a scene followed by a label and had to indicate whether the label matched the scene. We calculated an image’s “categorizability” score as the proportion of correct responses on congruent trials. These scores showed high levels of consistency, suggesting categorizability can be considered an intrinsic image property too. However, we did not observe a correlation between categorizability and memorability, suggesting the ease with which an image can be categorized relies on features distinct from those involved in memorability. Finally, my poster also presents plans for a study using information in the middle layers of CNNs to predict memorability.sponsorship: This work was supported by a personal fellowship awarded to Lore Goetschalckx by the Research Foundation - Flanders (FWO; Grant 1108116N) and by a Methusalem grant awarded to Johan Wagemans by the Flemish Government (METH/14/02).status: Publishe
MemCat: a new category-based image set quantified on memorability
No full textImages differ in their memorability in consistent ways across observers. What makes an image memorable is not fully understood to date. Most of the current insight is in terms of high-level semantic aspects, related to the content. However, research still shows consistent differences within semantic categories, suggesting a role for factors at other levels of processing in the visual hierarchy. To aid investigations into this role as well as contributions to the understanding of image memorability more generally, we present MemCat. MemCat is a category-based image set, consisting of 10K images representing five broader, memorability-relevant categories (animal, food, landscape, sports, and vehicle) and further divided into subcategories (e.g., bear). They were sampled from existing source image sets that offer bounding box annotations or more detailed segmentation masks. We collected memorability scores for all 10 K images, each score based on the responses of on average 99 participants in a repeat-detection memory task. Replicating previous research, the collected memorability scores show high levels of consistency across observers. Currently, MemCat is the second largest memorability image set and the largest offering a category-based structure. MemCat can be used to study the factors underlying the variability in image memorability, including the variability within semantic categories. In addition, it offers a new benchmark dataset for the automatic prediction of memorability scores (e.g., with convolutional neural networks). Finally, MemCat allows the study of neural and behavioral correlates of memorability while controlling for semantic category.sponsorship: This work was supported by a personal fellowship by the Research Foundation - Flanders (FWO) awarded to Lore Goetschalckx (Grant 1108116N), and by a Methusalem grant awarded to Johan Wagemans by the Flemish Government (METH/14/02). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. (Research Foundation - Flanders (FWO)|1108116N, Flemish Government|METH/14/02)status: Publishe
Image memorability across longer time intervals
No full textYou may find some images easier to remember than others. Recent studies of visual memory have found remarkable levels of consistency for this inter-item variability across observers, suggesting that memorability can be considered an intrinsic image property. The current study replicated and extended previous results, while adopting a more traditional visual longterm memory task with retention intervals of 20 min, one day, and one week, as opposed to
the previously used repeat-detection task, which typically relied on short retention intervals (5 min). Our memorability rank scores show levels of consistency across observers in line with those reported in previous research. They correlate strongly with previous quantifications and
appear stable over time. Furthermore, we show that the way consistency of memorability scores increases with the number of responses per image follows the Spearman–Brown
formula. Interestingly, our results also seem to show an increase in consistency with an increase in retention interval. Supported by simulated data, this effect is attributed to a decrease of extraneous influences on recognition over time. Finally, we also provide evidence
for a log-linear, rather than linear, decline of the raw memorability scores over time, with more memorable images declining less strongly.sponsorship: This work was supported by a personal fellowship awarded to Lore Goetschalckx by the Research Foundation - Flanders (Fonds Wetenschappelijk Onderzoek (FWO)) [grant number 1108116N] and by a Methusalem grant awarded to Johan Wagemans by the Flemish Government [METH/14/02]. (Research Foundation - Flanders (Fonds Wetenschappelijk Onderzoek (FWO))|1108116N, Flemish Government|METH/14/02)status: Publishe
Poster presentation: Are memorable images easier to categorize rapidly and do they survive shrinking better?
No full textSome images are intrinsically more memorable than others (e.g., Isola et al., 2014) but the features making them memorable are not yet well understood. Based on the hypothesis that memorability depends on perceptual organization, we present two studies in which we ask whether (i) more memorable images are also easier to categorize rapidly, and (ii) whether they survive shrinking to thumbnail size better.
We used meaningful real-life scene photographs from a previous memorability study (Bylinskii et al., 2015). In the rapid-categorization study, on each trial such an image (32 ms) was presented, directly followed by a mask (80 ms) and a category label. Participants indicated whether the label matched the image. An image's 'categorizability'-score was calculated as the proportion of correct responses on yes-trials. In the shrinking study, a regular-sized image, surrounded by nine thumbnails was presented on each trial. Participants had to locate the shrunken version of the middle image as fast as possible. An image's 'shrinkability'-score was the mean RT across correct responses.
Most categories showed high consistency (mean split-half Spearman's ρ up to .90) for both output variables, suggesting these are intrinsic image properties too. However, the predicted correlation with memorability was only observed for shrinkability (ρ = -.22), not for categorizability (ρ = -.05). To rule out distinctiveness as a confounding variable for the absence and/or presence of the observed correlations, we are currently quantifying the images on that variable. This additional analysis might provide a deeper insight into the current, unexpected pattern of results.sponsorship: PhD Fellowship of the Research Foundation - Flanders (FWO), granted to Lore Goetschalckx (Grant 1108116N). Methusalem grant awarded to Johan Wagemans by the Flemish Government (METH/14/02)status: Publishe
Increasing the Diversity in RGB-to-Thermal Image Translation for Automotive Applications
No full textMemCat: A new category-based image set quantified on memorability
No full texthttp://gestaltrevision.be/projects/memcat
MemCat: A new category-based image set quantified on memorability
No full textThis is a preprint. Please find the published, peer reviewed version of the paper here: https://peerj.com/articles/8169/. Images differ in their memorability in consistent ways across observers. What makes an image memorable is not fully understood to date. Most of the current insight is in terms of high-level semantic aspects, related to the content. However, research still shows consistent differences within semantic categories, suggesting a role for factors at other levels of processing in the visual hierarchy. To aid investigations into this role as well as contributions to the understanding of image memorability more generally, we present MemCat. MemCat is a category-based image set, consisting of 10K images representing five broader, memorability-relevant categories (animal, food, landscape, sports, and vehicle) and further divided into subcategories (e.g., bear). They were sampled from existing source image sets that offer bounding box annotations or more detailed segmentation masks. We collected memorability scores for all 10K images, each score based on the responses of on average 99 participants in a repeat-detection memory task. Replicating previous research, the collected memorability scores show high levels of consistency across observers. Currently, MemCat is the second largest memorability image set and the largest offering a category-based structure. MemCat can be used to study the factors underlying the variability in image memorability, including the variability within semantic categories. In addition, it offers a new benchmark dataset for the automatic prediction of memorability scores (e.g., with convolutional neural networks). Finally, MemCat allows to study neural and behavioral correlates of memorability while controlling for semantic category
MemCat: a new category-based image set quantified on memorability
No full textImages differ in their memorability in consistent ways across observers. What makes an image memorable is not fully understood to date. Most of the current insight is in terms of high-level semantic aspects, related to the content. However, research still shows consistent differences within semantic categories, suggesting a role for factors at other levels of processing in the visual hierarchy. To aid investigations into this role as well as contributions to the understanding of image memorability more generally, we present MemCat. MemCat is a category-based image set, consisting of 10K images representing five broader, memorability-relevant categories (animal, food, landscape, sports, and vehicle) and further divided into subcategories (e.g., bear). They were sampled from existing source image sets that offer bounding box annotations or more detailed segmentation masks. We collected memorability scores for all 10 K images, each score based on the responses of on average 99 participants in a repeat-detection memory task. Replicating previous research, the collected memorability scores show high levels of consistency across observers. Currently, MemCat is the second largest memorability image set and the largest offering a category-based structure. MemCat can be used to study the factors underlying the variability in image memorability, including the variability within semantic categories. In addition, it offers a new benchmark dataset for the automatic prediction of memorability scores (e.g., with convolutional neural networks). Finally, MemCat allows the study of neural and behavioral correlates of memorability while controlling for semantic category
Investigating the role of mid-level features in image memorability: A new image set
No full textstatus: Publishe
