PRESS-NEWS.org - Press Release Distribution
PRESS RELEASES DISTRIBUTION

Enhancing automatic image cropping models with advanced adversarial techniques

Researchers have developed techniques to create perturbations in images that can deceive cropping models, helping identify and address their weaknesses

Enhancing automatic image cropping models with advanced adversarial techniques
2024-08-01
(Press-News.org)

Image cropping is an essential task in many contexts, right from social media and e-commerce to advanced computer vision applications. Cropping helps maintain image quality by avoiding unnecessary resizing, which can degrade the image and consume computational resources. It is also useful when an image needs to conform to a predetermined aspect ratio, such as in thumbnails. Over the past decade, engineers around the world have developed various machine learning (ML) models to automatically crop images. These models aim to crop an input image in a way that preserves its most relevant parts.

However, these models can make mistakes and exhibit biases that, in the worst cases, can put users at legal risk. For example, in 2020, a lawsuit was filed against X (formerly Twitter) because its automatic cropping function hid the copyright information in a retweeted image. Therefore, it is crucial to understand the reason image cropping ML models fail so as to train and use them accordingly and avoid such problems.

Against this background, a research team from Doshisha University, Japan, set out to develop new techniques to generate adversarial examples for the task of image cropping. As explained in their latest paper, published in IEEE Access on June 17, 2024, their methods can introduce imperceptible noisy perturbations into an image to trick models into cropping regions that aligns with user intentions, even if the original model would have missed it. Doctoral student Masatomo Yoshida, the first author and lead researcher of the study, explains their motivation by saying, “To the best of our knowledge, there is very little research on adversarial attacks on image cropping models, as most previous research has focused on image classification and detection. These models need to be refined to ensure they respect user intentions and eliminate biases as much as possible while cropping images.”  Masatomo Yoshida and Haruto Namura from the Graduate School of Science and Engineering, Doshisha University, Kyoto, Japan and Masahiro Okuda is from the Faculty of Science and Engineering at Doshisha University were also a part of the study.

The researchers developed and implemented two distinct approaches for generating adversarial examples — a white-box approach and a black-box approach. The white-box method, requiring access to the internal workings of the target model, involves iteratively calculating perturbations to input images based on the model’s gradients. By employing a gaze prediction model to identify salient points within an image, this approach manipulates gaze saliency maps to achieve effective adversarial examples. It significantly reduces perturbation sizes, achieving a minimum size 62.5% smaller than baseline methods across an experimental image dataset.

The black-box approach utilizes Bayesian optimization to effectively narrow the search space and target specific image regions. Similar to the white-box strategy, this approach involves iterative procedures based on gaze saliency maps. Instead of using internal gradients, it employs a tree-structured Parzen estimator to select and optimize pixel coordinates that influence gaze saliency, ultimately producing desired adversarial images. Notably, black-box techniques are more broadly applicable in real-world scenarios and hold greater relevance in cybersecurity contexts.  

Both approaches show promise based on experimental outcomes, as graduate student Haruto Namura, a participant in the study, explains, “Our findings indicate that our methods not only surpass existing techniques but also show potential as effective solutions for real-world applications, such as those on platforms like Twitter.”

Overall, this study represents a significant advancement toward more reliable AI systems, crucial for meeting public expectations and earning their trust. Enhancing the efficiency of generating adversarial examples for image cropping will propel research in ML and inspire solutions to its pressing challenges. As Professor Masahiro Okuda, advisor to Namura and Yoshida, concludes, “By identifying vulnerabilities in increasingly deployed AI models, our research contributes to the development of fairer AI systems and addresses the growing need for AI governance.”

We will surely be keeping an eye out for further progress in this area!
 

About Masatomo Yoshida from Doshisha University, Japan
Masatomo Yoshida received his B.E. and M.E. degrees from Doshisha University, Kyoto, Japan, in 2021 and 2023, respectively. He is currently pursuing a Ph.D. degree in engineering at the Graduate School of Science and Engineering of Doshisha University. His research interests include analyzing spatio-temporal time series data, image processing, deep learning, and adversarial examples. He was awarded the Support for Pioneering Research Initiated by the Next Generation (SPRING) Scholarship by the Japan Science and Technology Agency (JST). Additionally, he received an award in a local photography contest held in Kyoto in 2015.

Funding information
This work was supported in part by the Japan Science and Technology Agency (JST) Support for Pioneering Research Initiated by the Next Generation (SPRING) under Grant JPMJSP2129, in part by the Japan Society for the Promotion of Science (JSPS) KAKENHI under Grant 23K11174, and in part by the Ministry of Education, Culture, Sports, Science and Technology (MEXT), Japan, Promotion of Distinctive Joint Research Center Program under Grant JPMXP 0621467946.

Media contact:
Organization for Research Initiatives & Development
Doshisha University
Kyotanabe, Kyoto 610-0394, JAPAN
E-mail:jt-ura@mail.doshisha.ac.jp

END


[Attachments] See images for this press release:
Enhancing automatic image cropping models with advanced adversarial techniques Enhancing automatic image cropping models with advanced adversarial techniques 2

ELSE PRESS RELEASES FROM THIS DATE:

$2.4 million grant helping MCG scientists better understand what happens to our skeleton as we age

$2.4 million grant helping MCG scientists better understand what happens to our skeleton as we age
2024-08-01
AUGUSTA, Ga. (Aug. 1, 2024) – Figuring out how the tissues in our bones, adrenal glands, muscle and fat “talk” to each other could help scientists better understand what happens to our skeletons with age, when our bones tend to lose mass and become weaker, leaving us at risk for falls and fractures. “Tissues don’t function in isolation – everything in the body “talks” to everything else to keep people healthy across the lifespan,” explains Meghan McGee-Lawrence, PhD, bone biologist at the Medical College of Georgia at ...

Using AI, USC researchers pioneer a potential new immunotherapy approach for treating glioblastoma

2024-08-01
In an innovative new study of glioblastoma, scientists used artificial intelligence (AI) to reprogram cancer cells, converting them into dendritic cells (DCs), which can identify cancer cells and direct other immune cells to kill them. Glioblastoma is the most common brain cancer in adults and also the deadliest, with less than 10% of patients surviving five years after their diagnosis. While new approaches such as immunotherapy have revolutionized treatment for other cancers, they have done little for patients with glioblastoma. That is partly because these hard-to-reach brain tumors ...

High blood pressure associated with environmental contamination by tellurium

High blood pressure associated with environmental contamination by tellurium
2024-08-01
The likelihood of developing high blood pressure (hypertension) increases with higher levels of tellurium, a contaminant transferred from mining and manufacturing activities to foods. Improved monitoring of tellurium levels in specific foods could help decrease high blood pressure in the general population. The results of a study examining the relationship between tellurium exposure and hypertension were published in the journal Environment International.    The study was led by Nagoya University in Japan. According to Takumi Kagawa, one of the researchers involved in the ...

Pursuing the middle path to scientific discovery

Pursuing the middle path to scientific discovery
2024-08-01
In electronic technologies, key material properties change in response to stimuli like voltage or current. Scientists aim to understand these changes in terms of the material’s structure at the nanoscale (a few atoms) and microscale (the thickness of a piece of paper). Often neglected is the realm between, the mesoscale — spanning 10 billionths to 1 millionth of a meter. Scientists at the U.S. Department of Energy’s (DOE) Argonne National Laboratory, in collaboration with Rice University and DOE’s Lawrence ...

Salk awarded $3.6 million by the California Institute for Regenerative Medicine to advance research on brain aging

Salk awarded $3.6 million by the California Institute for Regenerative Medicine to advance research on brain aging
2024-08-01
LA JOLLA (July 31, 2024)—The Salk Institute was awarded $3.6 million by the California Institute for Regenerative Medicine (CIRM), one of the world’s largest institutions dedicated to regenerative medicine. Salk Professor Rusty Gage will lead the new CIRM-funded Shared Resources Laboratory focused on stem cell-based models of aging and neurodegeneration. The award is part of CIRM’s latest round of funding to address challenges in the regenerative medicine field. The state agency dedicated $27 million to help establish six new Shared Resources ...

Generation X and millennials in US have higher risk of developing 17 cancers compared to older generations, new study suggests

Generation X and millennials in US have higher risk of developing 17 cancers compared to older generations, new study suggests
2024-08-01
A new large study led by researchers at the American Cancer Society (ACS) suggests incidence rates continued to rise in successively younger generations in 17 of the 34 cancer types, including breast, pancreatic, and gastric cancers. Mortality trends also increased in conjunction with the incidence of liver (female only), uterine corpus, gallbladder, testicular, and colorectal cancers. The report will be published today in the journal The Lancet Public Health. “These findings add to growing evidence of increased cancer risk in post-Baby Boomer ...

Around 160,000 joint replacement surgeries lost by COVID-19 pandemic, study finds

2024-08-01
Nearly nine months of joint replacement surgery has been lost - around 160,000 fewer operations –  since the start of the COVID-19 pandemic, a new study led by the University of Bristol has found.  The research suggests returning to pre-pandemic levels will not tackle the backlog, and even with rapid expansion, it will take many years, if not decades, to fix this joint replacement crisis. The study, published in The Bone & Joint Journal today [1 August], looked in detail at the effect ...

Public health measures that reduce dementia risk could save up to £4bn

2024-07-31
Public health interventions that tackle dementia risk factors could yield as much as £4bn in savings in England by reducing dementia rates and helping people to live longer and healthier, finds a new study led by UCL researchers. The study, published in The Lancet Healthy Longevity, shows that interventions – such as reformulating food products to reduce sugar and salt intake, introducing low emission zones to improve air quality in cities, and minimum alcohol unit pricing to reduce drinking – could have extensive benefits beyond just the health outcomes they are directly targeting. Lead author Dr Naaheed Mukadam (UCL Psychiatry) ...

The Lancet: Nearly half of dementia cases could be prevented or delayed by tackling 14 risk factors starting in childhood, including two new risks—high cholesterol and vision loss

2024-07-31
Vision loss and high cholesterol add to 12 previously identified modifiable risk factors for dementia, concludes a new report from the 2024 Lancet Commission. The potential to prevent and better manage dementia is high if action to tackle these risk factors begins in childhood and continues throughout life, even in individuals with high genetic risk for dementia. New report outlines 13 recommendations for individuals and governments to help reduce risk, including preventing and treating hearing loss, vision loss, and depression; being cognitively active ...

Precision oncology via artificial intelligence on cancer biopsies

Precision oncology via artificial intelligence on cancer biopsies
2024-07-31
A new generation of artificial intelligence (AI) tools designed to allow rapid, low-cost detection of clinically actionable genomic alterations directly from tumor biopsy slides has been developed by a team led by engineers and medical researchers at University of California San Diego. A paper describing the new AI protocol for examining routine biopsies, called DeepHRD, was recently published in the Journal of Clinical Oncology. Senior author Ludmil Alexandrov, Ph.D., professor of bioengineering and professor of cellular and molecular medicine at UC San Diego, says the new method is designed to save weeks and thousands of dollars from clinical oncology treatment workflows ...

LAST 30 PRESS RELEASES:

Thirty-year mystery of dissonance in the “ringing” of black holes explained

Less intensive works best for agricultural soil

Arctic rivers project receives “national champion” designation from frontiers foundation

Computational biology paves the way for new ALS tests

Study offers new hope for babies born with opioid withdrawal syndrome

UT, Volkswagen Group of America celebrate research partnership

New Medicare program could dramatically improve affordability for cancer drugs – if patients enroll

Are ‘zombie’ skin cells harmful or helpful? The answer may be in their shapes

University of Cincinnati Cancer Center presents research at AACR 2025

Head and neck, breast, lung and survivorship studies headline Dana-Farber research at AACR Annual Meeting 2025

AACR: Researchers share promising results from MD Anderson clinical trials

New research explains why our waistlines expand in middle age

Advancements in muon detection: Taishan Antineutrino Observatory's innovative top veto tracker

Chips off the old block

Microvascular decompression combined with nerve combing for atypical trigeminal neuralgia

Cutting the complexity from digital carpentry

Lung immune cell type “quietly” controls inflammation in COVID-19

Fiscal impact of expanded Medicare coverage for GLP-1 receptor agonists to treat obesity

State and sociodemographic trends in US cigarette smoking with future projections

Young adults drive historic decline in smoking

NFCR congratulates Dr. Robert C. Bast, Jr. on receiving the AACR-Daniel D. Von Hoff Award for Outstanding Contributions to Education and Training in Cancer Research

Chimpanzee stem cells offer new insights into early embryonic development

This injected protein-like polymer helps tissues heal after a heart attack

FlexTech inaugural issue launches, pioneering interdisciplinary innovation in flexible technology

In Down syndrome mice, 40Hz light and sound improve cognition, neurogenesis, connectivity

Methyl eugenol: potential to inhibit oxidative stress, address related diseases, and its toxicological effects

A vascularized multilayer chip reveals shear stress-induced angiogenesis in diverse fluid conditions

AI helps unravel a cause of Alzheimer's disease and identify a therapeutic candidate

Coalition of Autism Scientists critiques US Department of Health and Human Services Autism Research Initiative

Structure dictates effectiveness, safety in nanomedicine

[Press-News.org] Enhancing automatic image cropping models with advanced adversarial techniques
Researchers have developed techniques to create perturbations in images that can deceive cropping models, helping identify and address their weaknesses