16,299 results on '"Isola"'
Search Results
2. The Pact
- Author
-
Isola, Dinamarie
- Published
- 2023
3. Work Ethics Practices and Employee Retention Towards Economic Recovery in the Brewery Industries of Lagos Metropolis, Nigeria
- Author
-
Babalola O. Oginni, Isola O. Ayantunji, Folakemi O. Larnre-Babalola, Zacchaeus O. Olonade, Kolawole S. Ajibola, Ramat A. Balogun, and Waseeu G. Adebayo
- Subjects
work ethics practices ,retention ,professionalism ,integrity ,accountability ,Political institutions and public administration (General) ,JF20-2112 ,Social sciences (General) ,H1-99 - Abstract
The study investigated work ethics practices and employee retention in the manufacturing industries in Lagos, Nigeria with focus on the Nigerian Breweries, Lagos as the unit of analysis. The study identified work ethics practices found operational to include integrity accountability, professionalism, and respectfulness. The corresponding relationship between these work ethics practices and employee retention was also evaluated with the view to providing information on how work ethics practices influence employee retention. The study made use of primary data with the aid of a structured questionnaire administered to 300 respondents randomly to elicit information from the selected respondents. It was found that integrity as a practice was prevalent among other practices found to be operational in the area of study and also found that there exists a significant positive relationship among all the variables of work ethics practices and employee retention where accountability practice with employee retention (r = 0.546, p 0.05), respectfulness practice with employee retention (r = 0.677, p 0.05), professionalism practice with employee retention (r = 0.658, p 0.05) and integrity practice with employee retention (r = 0.748, p 0.05). it was concluded that work ethics practice is a good cost control strategy to sustain employee retention without incurring additional costs, especially in the wake of economic recession and recovery. Thus, recommended that the management of the Nigerian Breweries should institutionalise work ethics practices and educate employees of the organisation on the essence of the practices.
- Published
- 2023
- Full Text
- View/download PDF
4. OCCUPATIONAL STRESS AND INDUSTRIAL RELATIONS OUTCOMES: EVIDENCE FROM SELECTED MANUFACTURING ORGANIZATIONS IN LAGOS METROPOLIS, NIGERIA
- Author
-
Babalola Oluwayemi Oginni, Olusegun Abel Adesanya, Gbenga Ezekiel Ilori, Isola Olalekan Ayantunji, Folakemi Olubunmi Lanre-Babalola, and Kolawole Sunday Ajibola
- Subjects
occupational stress ,physiological stress ,psychological stress ,industrial relations outcomes ,absenteeism. ,Business ,HF5001-6182 - Abstract
Purpose: The study investigated the relationship between occupational stress and industrial relations outcomes using the manufacturing organizations in Lagos Metropolis as the unit of analysis. The study identified two dimensions of occupational stress (psychological and physiological stress) and four industrial relations outcomes (quit, absenteeism, industrial actions, and management-employee relations). Methodology: The study made use of multi-sampling techniques for the preservation of the characteristics and status of the respondents and the research questionnaire was administered randomly to the selected sample size of 363 respondents. Findings and implications: It was found that there exists a positive and significant relationship between and among all the variables of occupational stress and industrial relations outcomes with absenteeism as the potent force. It was concluded that occupational stress is dominant and permanent in the selected manufacturing organizations and manifested in the form of absenteeism. Limitations: The study concentrated on the Lagos Metropolis, but further studies can look farther into other geopolitical zones in Nigeria to further expand the frontier of knowledge in the area of study. Originality: The study provides a more comprehensive understanding of occupational stress dimensions on industrial relations outcomes, thus, expanding the existing literature in the area of study.
- Published
- 2023
- Full Text
- View/download PDF
5. Effect of sawdust mulch on soil properties and performance of tomatoes (Lycopersicum esculentum) in an alfisol in Southwestern Nigeria
- Author
-
Femi Olusola Abiodun and Isola John Oluwasina
- Subjects
tomato ,sawdust ,mulching ,soil ,Agriculture - Abstract
Mulching is an effective method of manipulating crop growing environment to increase yield. A field experiment was carried out at the Teaching and Research Farm of the Federal College of Forestry Jericho (Ibadan). The study aimed to examine the effect of sawdust mulching on selected soil properties, growth and yield of two tomato varieties. The study was a 4 x 2 factorial experiment laid out in a randomized completely block design (RCBD) with four replications. Factor one consisted of four treatments, namely: T0 (control), T1 (sawdust mulch [SDM] at 10 t ha-1 ), T2 (SDM at 20 t ha -1 ) and T3 (SDM at 30 t ha1 ). The second factor was the two tomato varieties (UC82B and Ibadan local). Plant height, stem girth, number of leaves and branches were collected at a 2-week interval while the number of fruits, fruit weight (kg/ha) and yield parameters were measured at maturity. The Genstat statistical software package was used for data analysis and an LSD test was performed at the 5% level of significance. The sawdust mulch used had a low nitrogen content (0.60%) and a moderate organic carbon content (38.6%). The use of sawdust mulch had a major impact on tomato yield, while there was not any significant variation among the examined tomato varieties. UC82B (117.92 kg ha-1 ) at 30 t ha-1 SDM plot had the highest yield, followed by Ibadan local (103.93 kg ha-1 ) at 30 t ha-1 SDM plot, while Ibadan local (61.94 kg ha-1 ) at the control plot had the lowest yield. In this study, the tomato performed best with sawdust mulch at 30 t ha-1 and is therefore recommended to farmers in the study area to maximize tomato production.
- Published
- 2023
- Full Text
- View/download PDF
6. THE NEXUS BETWEEN RECRUITMENT AND SELECTION PRACTICES AND BUSINESS PERFORMANCE OF NIGERIAN SMALL AND MEDIUM-SIZED MANUFACTURING ENTERPRISES
- Author
-
Babalola Oluwayemi Oginni, Isola Olalekan Ayantunji, Toyin Solomon Olaniyan, Kolawole Sunday Ajibola, and Theresa ‘Bosede Ajakaye
- Subjects
Management. Industrial management ,HD28-70 ,Marketing. Distribution of products ,HF5410-5417.5 - Abstract
This paper examined the relationships between recruitment and selection practices with business performance in the manufacturing sector using the Small and Medium Enterprises (SMEs) in the Lagos metropolis as the unit of analysis. It identified employee referral, job vacancy advertisement, employment agency and employee promotion as the most commonly used recruitment and selection practices in the manufacturing SMEs’ environment. The study made use of a questionnaire administered to 288 respondents and the results obtained showed that the vacancy advertisement and employment agency have a moderate relationship with business performance while employee promotion has a weak relationship with business performance while employee referral has a negative relationship with business performance. Regression results also indicated a positive relationship between combined recruitment and selection practices and business performance. It was concluded that the management of SMEs should prioritise employment agency among the recruitment and selection practices since it was found to have a more decisive influence on business performance. Thus, having the patronage of the employment agency with sound personnel policies that will be costeffective and outline human resource requirements is recommended.
- Published
- 2023
- Full Text
- View/download PDF
7. Nexus between entrepreneurial characteristics and small business productivity in Nigeria
- Author
-
Babalola O. Oginni, Samuel O. Omoyele, Isola O. Ayantunji, Folakemi O. Larnre-Babalola, and Ramat A. Balogun
- Subjects
entrepreneurs’ characteristics ,business performance ,innovativeness ,orientation ,risk-taking propensity ,experience ,Finance ,HG1-9999 ,Economics as a science ,HB71-74 - Abstract
Purpose — The productivity challenge confronting small businesses in developing countries has been identified among others to include entrepreneurs’ characteristics. The objective of the study was to investigate the effects of entrepreneurs’ characteristics on the business productivity of SMEs in Nigeria and anchored on the entrepreneur’s innovativeness, experience, orientation, and risk-taking propensity. Method — It was a cross-sectional study carried out among entrepreneurs of SMEs in Southwest Nigeria using a descriptive research survey design. Respondents were selected using a convenience sampling technique. A sample size of 400 respondents was selected for the study, and descriptive statistics such as frequencies and percentages were adopted for data analysis. The hypotheses formulated were tested using chi-square at 0.05 level of significance. Result — The result of the study revealed that an entrepreneur’s characteristics, such as innovativeness, experiences, orientation, and risk-taking propensity, significantly impact productivity. However, innovativeness and risk-taking were paramount among other entrepreneurial characteristics of SMEs in Nigeria. Contribution — The result validated the works of the earlier scholars in the study area and contributed to expanding literature on how entrepreneurs’ characteristics, especially their innovativeness, experience, orientation, and desire to take risks, can solve small business challenges in developing economies.
- Published
- 2023
- Full Text
- View/download PDF
8. Adaptive Length Image Tokenization via Recurrent Allocation
- Author
-
Duggal, Shivam, Isola, Phillip, Torralba, Antonio, and Freeman, William T.
- Subjects
Computer Science - Computer Vision and Pattern Recognition ,Computer Science - Artificial Intelligence ,Computer Science - Machine Learning ,Computer Science - Robotics - Abstract
Current vision systems typically assign fixed-length representations to images, regardless of the information content. This contrasts with human intelligence - and even large language models - which allocate varying representational capacities based on entropy, context and familiarity. Inspired by this, we propose an approach to learn variable-length token representations for 2D images. Our encoder-decoder architecture recursively processes 2D image tokens, distilling them into 1D latent tokens over multiple iterations of recurrent rollouts. Each iteration refines the 2D tokens, updates the existing 1D latent tokens, and adaptively increases representational capacity by adding new tokens. This enables compression of images into a variable number of tokens, ranging from 32 to 256. We validate our tokenizer using reconstruction loss and FID metrics, demonstrating that token count aligns with image entropy, familiarity and downstream task requirements. Recurrent token processing with increasing representational capacity in each iteration shows signs of token specialization, revealing potential for object / part discovery., Comment: Code at: https://github.com/ShivamDuggal4/adaptive-length-tokenizer
- Published
- 2024
9. Learning Visual Parkour from Generated Images
- Author
-
Yu, Alan, Yang, Ge, Choi, Ran, Ravan, Yajvan, Leonard, John, and Isola, Phillip
- Subjects
Computer Science - Robotics - Abstract
Fast and accurate physics simulation is an essential component of robot learning, where robots can explore failure scenarios that are difficult to produce in the real world and learn from unlimited on-policy data. Yet, it remains challenging to incorporate RGB-color perception into the sim-to-real pipeline that matches the real world in its richness and realism. In this work, we train a robot dog in simulation for visual parkour. We propose a way to use generative models to synthesize diverse and physically accurate image sequences of the scene from the robot's ego-centric perspective. We present demonstrations of zero-shot transfer to the RGB-only observations of the real world on a robot equipped with a low-cost, off-the-shelf color camera. website visit https://lucidsim.github.io, Comment: 17 pages, 19 figures
- Published
- 2024
10. When Does Perceptual Alignment Benefit Vision Representations?
- Author
-
Sundaram, Shobhita, Fu, Stephanie, Muttenthaler, Lukas, Tamir, Netanel Y., Chai, Lucy, Kornblith, Simon, Darrell, Trevor, and Isola, Phillip
- Subjects
Computer Science - Computer Vision and Pattern Recognition ,Computer Science - Machine Learning - Abstract
Humans judge perceptual similarity according to diverse visual attributes, including scene layout, subject location, and camera pose. Existing vision models understand a wide range of semantic abstractions but improperly weigh these attributes and thus make inferences misaligned with human perception. While vision representations have previously benefited from alignment in contexts like image generation, the utility of perceptually aligned representations in more general-purpose settings remains unclear. Here, we investigate how aligning vision model representations to human perceptual judgments impacts their usability across diverse computer vision tasks. We finetune state-of-the-art models on human similarity judgments for image triplets and evaluate them across standard vision benchmarks. We find that aligning models to perceptual judgments yields representations that improve upon the original backbones across many downstream tasks, including counting, segmentation, depth estimation, instance retrieval, and retrieval-augmented generation. In addition, we find that performance is widely preserved on other tasks, including specialized out-of-distribution domains such as in medical imaging and 3D environment frames. Our results suggest that injecting an inductive bias about human perceptual knowledge into vision models can contribute to better representations., Comment: S.S. and S.F. contributed equally. Website: percep-align.github.io
- Published
- 2024
11. Dietary Factors Affecting the Prevalence and Impact of Periodontal Disease
- Author
-
Santonocito S, Polizzi A, Palazzo G, Indelicato F, and Isola G
- Subjects
periodontitis ,nutrition ,diet ,macronutrients ,micronutrients ,oral health ,Dentistry ,RK1-715 - Abstract
Simona Santonocito, Alessandro Polizzi, Giuseppe Palazzo, Francesco Indelicato, Gaetano Isola Department of General Surgery and Surgical-Medical Specialties, School of Dentistry, University of Catania, Catania, ItalyCorrespondence: Gaetano IsolaUnit of Periodontology, Department of General Surgery and Surgical-Medical Specialties, School of Dentistry, University of Catania, Via S. Sofia 78, Catania, 95123, ItalyTel/Fax +390953782453Email gaetano.isola@unict.itAbstract: In the last few decades, growing evidence have shown a possible impact of diet and nutrients on oral health. This review aims to describe, in the light of current knowledge, the role of diet, nutrients, and micronutrients in periodontal health and periodontal diseases. A variety of macronutrients and micronutrients could have an impact on periodontal health. The balanced intake of unprocessed complex carbohydrates, vegetable proteins, omega-3 fatty acids, minerals, and vitamins positively affects periodontal inflammation. On the other way, refined carbohydrates, non-vegetable proteins, proinflammatory saturated fatty acids and an unbalanced supply of vitamins and minerals may increase periodontal inflammation. This review will discuss the current evidence that shows how a healthy and balanced diet has anti-inflammatory and protective effects on periodontal health. Therefore, it appears that adopting a correct lifestyle and diet should be encouraged in patients with oral and periodontal disease.Keywords: periodontitis, nutrition, diet, macronutrients, micronutrients, oral health
- Published
- 2021
12. Self-Similar Topological Fractals
- Author
-
Cipriani, Fabio E. G., Guido, Daniele, Isola, Tommaso, and Sauvageot, Jean-Luc
- Subjects
Mathematics - Functional Analysis ,28A80 - Abstract
We introduce the notion of (abelian) similarity scheme, as a constructive model for topological self-similar fractals, in the same way in which the notion of iterated function system furnishes a constructive notion of self-similar fractals in a metric environment. At the same time, our notion gives a constructive approach to the Kigami-Kameyama notion of topological fractals, since a similarity scheme produces a topological fractal a la Kigami-Kameyama, and many Kigami-Kameyama topological fractals may be constructed via similarity schemes. Our scheme consists of objects $X_0\stackrel{\varphi}{\rightarrow}X_1\stackrel{\pi}{\leftarrow} Y\times X_0$, where $X_0,X_1$ and $Y$ are compact Hausdorff spaces, the map $\varphi$ is continuous injective and the map $\pi$ is continuous surjective. This scheme produces a sequence $X_n$, $n\in\mathbb{N}$, of compact Hausdorff spaces, $X_n$ embedded in $X_{n+1}$, and a compact Hausdorff space $X_\infty$ giving a sort of injective limit space, which turns out to be self-similar. We observe that the space $Y$ parametrizes the generalized similarity maps, and finiteness of $Y$ is not required., Comment: 20 pages
- Published
- 2024
13. Community participation in the provision of environmental sanitation infrastructure in Akure, Nigeria
- Author
-
Olugbamila Omotayo B., Adeyinka Samson A., Odunsi Oluwafemi M., Olowoyo Sanya A., Isola Oluwadara L., and Adanlawo Tolulope D.
- Subjects
sanitation exercise ,citizen participation ,environmental facilities ,environmental amenities ,urbanisation ,urban planning ,Environmental sciences ,GE1-350 - Abstract
This article assessed community participation in the provision of environmental sanitation infrastructure in Akure, Nigeria. A multi-stage sampling procedure was employed for the study. Four residential zones: the traditional core, the transition zone, the peripheral zone and the public housing district were identified, this was with a view to showcasing the variation in the level of contributions and efforts of different communities based on the delineated areas towards the provision of environmental sanitation infrastructure in the study area. Twenty residential areas were identified across the residential zones of the study area out of which ten areas which represented 50% of the total areas were selected. Ten percent of the total number of buildings in the selected areas were subsequently sampled, resulting in the selection of 180 residents for questionnaire administration. The data collected were analysed using descriptive statistics with frequency used for univariate analysis and cross tabulation for bivariate and multivariate analysis as well as the use of chi-square for inferential statistics. Findings revealed that community participation did not play a leading role in providing environmental sanitation facilities in Akure, but that the government did. This shows that most of the environmental facilities in the area are provided by the government, thus revealing the overdependence of the residents on the government for the provision of environmental sanitation infrastructure. Further findings revealed that most of the challenges faced in the study area in terms of providing environmental sanitation infrastructure had a significant influence on the provision of these facilities. Thus, the study showed that challenges significantly hindered the provision of environmental sanitation facilities in the area.
- Published
- 2020
- Full Text
- View/download PDF
14. A New Methodology for the Digital Planning of Micro-Implant-Supported Maxillary Skeletal Expansion
- Author
-
Cantarella D, Savio G, Grigolato L, Zanata P, Berveglieri C, Lo Giudice A, Isola G, Del Fabbro M, and Moon W
- Subjects
miniscrew-assisted rapid palatal expansion (marpe) ,cone beam computed tomography (cbct) ,mse ,digital planning and workflow ,virtual planning ,temporary anchorage device (tad) ,cephalometrics-based digital planning (cbdp) ,Medical technology ,R855-855.5 - Abstract
Daniele Cantarella,1 Gianpaolo Savio,2 Luca Grigolato,2 Paolo Zanata,3 Chiara Berveglieri,4 Antonino Lo Giudice,5,6 Gaetano Isola,5 Massimo Del Fabbro,1 Won Moon7 1Department of Biomedical, Surgical and Dental Sciences, University of Milan, Milan, Italy; 2Department of Civil, Environmental and Architectural Engineering ICEA, University of Padova, Padova, Italy; 3Private Practice, Castelfranco Veneto, Italy; 4Private Practice of Orthodontics, Bondeno, Italy; 5Department of General Surgery and Surgical-Medical Specialties, Section of Orthodontics, School of Dentistry, University of Catania, Catania, Italy; 6Department of Biomedical and Dental Sciences and Morphofunctional Imaging, Section of Orthodontics, University of Messina, Messina, Italy; 7Division of Growth and Development, Section of Orthodontics, School of Dentistry, Center for Health Science, University of California, Los Angeles, Los Angeles, CA, USACorrespondence: Daniele CantarellaDepartment of Biomedical, Surgical and Dental Sciences, University of Milan, Via Commenda 10, Milan, ItalyEmail danielecant@hotmail.comIntroduction: Miniscrew-assisted rapid palatal expansion (MARPE) appliances utilize the skeletal anchorage to expand the maxilla. One type of MARPE device is the Maxillary Skeletal Expander (MSE), which presents four micro-implants with bicortical engagement of the palatal vault and nasal floor. MSE positioning is traditionally planned using dental stone models and 2D headfilms. This approach presents some critical issues, such as the inability to identify the MSE position relative to skeletal structures, and the potential risk of damaging anatomical structures.Methods: A novel methodology has been developed to plan MSE position using the digital model of dental arches and cone-beam computed tomography (CBCT). A virtual model of MSE appliance with the four micro-implants was created. After virtual planning, a positioning guide is virtually designed, 3D printed, and utilized to model and weld the MSE supporting arms to the molar bands. The expansion device is then cemented in the patient oral cavity and micro-implants inserted. A clinical case of a 12.9-year-old female patient presenting a Class III malocclusion with transverse and sagittal maxillary deficiency is reported.Results: The midpalatal suture was opened with a split of 3.06 mm and 2.8 mm at the anterior and posterior nasal spine, respectively. After facemask therapy, the sagittal skeletal relationship was improved, as shown by the increase in ANB, A-Na perpendicular and Wits cephalometric parameters, and the mandibular plane rotated 1.6° clockwise.Conclusion: The proposed digital methodology represents an advancement in the planning of MSE positioning, compared to the traditional approach. By evaluating the bone morphology of the palate and midface on patient CBCT, the placement of MSE is improved regarding the biomechanics of maxillary expansion and the bone thickness at micro-implants insertion sites. In the present case report, the digital planning was associated with a positive outcome of maxillary expansion and protraction in safety conditions.Keywords: miniscrew-assisted rapid palatal expansion, MARPE, CBCT, MSE, virtual planning, cephalometrics-based digital planning, CBDP, workflow, TAD
- Published
- 2020
15. P-032: INTEGRATIVE DIAGNOSIS OF SICKLE CELL DISEASE PATIENTS FOR PERSONALIZED MEDICINE
- Author
-
IDRIZOVIC A., COLLADO GIMBERT A., VAN BEERS E., COLOMBATTI R., BARTOLUCCI P., DE MONTALEMBERT M., BOARO M., BENEITEZ D., ORTUÑO A., RUIZ A., ISOLA I., CELA E., VAN WIJK R., RAB M., and MAÑU PEREIRA M.
- Subjects
Diseases of the blood and blood-forming organs ,RC633-647.5 - Published
- 2022
- Full Text
- View/download PDF
16. Scalable Optimization in the Modular Norm
- Author
-
Large, Tim, Liu, Yang, Huh, Minyoung, Bahng, Hyojin, Isola, Phillip, and Bernstein, Jeremy
- Subjects
Computer Science - Machine Learning - Abstract
To improve performance in contemporary deep learning, one is interested in scaling up the neural network in terms of both the number and the size of the layers. When ramping up the width of a single layer, graceful scaling of training has been linked to the need to normalize the weights and their updates in the "natural norm" particular to that layer. In this paper, we significantly generalize this idea by defining the modular norm, which is the natural norm on the full weight space of any neural network architecture. The modular norm is defined recursively in tandem with the network architecture itself. We show that the modular norm has several promising applications. On the practical side, the modular norm can be used to normalize the updates of any base optimizer so that the learning rate becomes transferable across width and depth. This means that the user does not need to compute optimizer-specific scale factors in order to scale training. On the theoretical side, we show that for any neural network built from "well-behaved" atomic modules, the gradient of the network is Lipschitz-continuous in the modular norm, with the Lipschitz constant admitting a simple recursive formula. This characterization opens the door to porting standard ideas in optimization theory over to deep learning. We have created a Python package called Modula that automatically normalizes weight updates in the modular norm of the architecture. The package is available via "pip install modula" with source code at https://github.com/jxbz/modula.
- Published
- 2024
17. The Platonic Representation Hypothesis
- Author
-
Huh, Minyoung, Cheung, Brian, Wang, Tongzhou, and Isola, Phillip
- Subjects
Computer Science - Machine Learning ,Computer Science - Artificial Intelligence ,Computer Science - Computer Vision and Pattern Recognition ,Computer Science - Neural and Evolutionary Computing - Abstract
We argue that representations in AI models, particularly deep networks, are converging. First, we survey many examples of convergence in the literature: over time and across multiple domains, the ways by which different neural networks represent data are becoming more aligned. Next, we demonstrate convergence across data modalities: as vision models and language models get larger, they measure distance between datapoints in a more and more alike way. We hypothesize that this convergence is driving toward a shared statistical model of reality, akin to Plato's concept of an ideal reality. We term such a representation the platonic representation and discuss several possible selective pressures toward it. Finally, we discuss the implications of these trends, their limitations, and counterexamples to our analysis., Comment: Equal contributions. Project: https://phillipi.github.io/prh/ Code: https://github.com/minyoungg/platonic-rep
- Published
- 2024
18. Editorial: Research Collaboration and Networks: Characteristics, Evolution and Trends
- Author
-
Isola Ajiferuke, Maria Cláudia Cabrini Grácio, and Siluo Yang
- Subjects
scientific collaboration ,research networks ,collaboration characteristics ,research collaboration methodology ,collaboration trends ,Bibliography. Library science. Information resources - Published
- 2021
- Full Text
- View/download PDF
19. Phenotypic plasticity of shell shape and growth of mussels to environmental conditions: Mytella strigata in the southeast Gulf of California
- Author
-
Gongora-Gomez, Andres Martin, Hernandez-Sepulveda, Juan Antonio, Isola, Tomas E., Sepulveda, Carlos Humberto, Montoya-Ponce, Celeste Osiris, and Garcia-Ulloa, Manuel
- Published
- 2024
- Full Text
- View/download PDF
20. Canagliflozin treatment prevents follicular exhaustion and attenuates hallmarks of ovarian aging in genetically heterogenous mice
- Author
-
Isola, José V. V., Biswas, Subhasri, Jayarathne, Hashan, Hubbart, Chase R., Hense, Jessica D., Matsuzaki, Satoshi, Kinter, Michael T., Humphries, Kenneth M., Ocañas, Sarah R., Sadagurski, Marianna, and Stout, Michael B.
- Published
- 2024
- Full Text
- View/download PDF
21. Influence of a Tensile Stress on the Thermodynamics of Martensitic Transformations: Frictional Work and Stored Elastic Energy Calculations in Polycrystalline Alloys
- Author
-
Malvasio, Bruno F., Isola, Lucio M., Giordana, M. Florencia, and Malarria, Jorge A.
- Published
- 2024
- Full Text
- View/download PDF
22. Case-by-case combination of the prostate imaging reporting and data system version 2.1 with the Likert score to reduce the false-positives of prostate MRI: a proof-of-concept study
- Author
-
Girometti, Rossano, Peruzzi, Valeria, Polizzi, Paolo, De Martino, Maria, Cereser, Lorenzo, Casarotto, Letizia, Pizzolitto, Stefano, Isola, Miriam, Crestani, Alessandro, Giannarini, Gianluca, and Zuiani, Chiara
- Published
- 2024
- Full Text
- View/download PDF
23. Lutein protection against doxorubicin-induced liver damage in rats is associated with inhibition of oxido-inflammatory stress and modulation of Beclin-1/mTOR activities
- Author
-
Asiwe, Jerome Ndudi, Yovwin, Godwin D., Alawode, Mercy Oluwalani, Isola, Theodora, Umukoro, Emuesiri Kohworho, Igbokwe, Vincent Ugochukwu, and Asiwe, Nicholas
- Published
- 2024
- Full Text
- View/download PDF
24. Risk factors for infection after carbapenem-resistant Acinetobacter baumannii colonization
- Author
-
Peghin, Maddalena, Givone, Filippo, de Martino, Maria, Ali, Raja Waqar, Graziano, Elena, Isola, Miriam, and Grossi, Paolo Antonio
- Published
- 2024
- Full Text
- View/download PDF
25. Faster steroid-free remission with tocilizumab compared to methotrexate in giant cell arteritis: a real-life experience in two reference centres
- Author
-
Quartuccio, Luca, Treppo, Elena, De Martino, Maria, Pillon, Maria, Perniola, Simone, Bruno, Dario, Isola, Miriam, and Gremese, Elisa
- Published
- 2024
- Full Text
- View/download PDF
26. Surface nanocavitation of titanium modulates macrophage activity
- Author
-
Ariganello MB, Guadarrama Bello D, Rodriguez-Contreras A, Sadeghi S, Isola G, Variola F, and Nanci A
- Subjects
Nanotopography ,Inflammation ,Osteopontin ,SPARC ,Stabilin-1 ,Cytokines ,Medicine (General) ,R5-920 - Abstract
Marianne B Ariganello,1,* Dainelys Guadarrama Bello,1,* Alejandra Rodriguez-Contreras,1 Shayan Sadeghi,1 Gaetano Isola,1 Fabio Variola,2,3 Antonio Nanci1,4 1Laboratory for the Study of Calcified Tissues and Biomaterials, Department of Stomatology, Faculty of Dental Medicine, Université de Montréal, Montreal, QC, Canada; 2Department of Mechanical Engineering, University of Ottawa, Ottawa, ON, Canada; 3Department of Cellular and Molecular Medicine, University of Ottawa, Ottawa, ON, Canada; 4Department of Biochemistry and Molecular Medicine, Faculty of Medicine, Université de Montréal, Montreal, QC, Canada *These authors contributed equally to this work Background: Nanoscale surface modifications are widely touted to improve the biocompatibility of medically relevant materials. Immune cells, such as macrophages, play a critical role in the initial healing events following implantation. Methods: To understand the response of macrophages to nanotopography better, we exposed U937-derived macrophages to a distinctive mesoporous titanium surface (TiNano) produced by a process of simple chemical nanocavitation, and to mechanically polished titanium (TiPolished) and glass coverslip (Glass) surfaces as controls. Cell numbers and morphology were evaluated. Osteopontin expression and that of the proinflammatory SPARC protein and its stabilin 1 receptor were analyzed. Release of inflammation-associated cytokines and chemokines was also measured. Results: Compared to the two control surfaces, there were fewer U937 cells on TiNano, and these exhibited a more rounded morphology with long filopodia. The cells showed areas of punctate actin distribution, indicating formation of podosomes. Of the three proteins examined, only osteopontin’s immunofluorescence signal was clearly reduced. Irrespective of the substrate, the cytokine assay revealed important variations in expression levels of the multiple molecules analyzed and downregulation in a number of chemokines by the TiNano surface. Conclusion: These results indicate that macrophages sense and respond to the physicochemical cueing generated by the nanocavitated surface, triggering cellular and molecular changes consistent with lesser inflammatory propensity. Given the previously reported beneficial outcome of this mesoporous surface on osteogenic activity, it could be presumed that modulation of the macrophagic response it elicits may also contribute to initial bone-integration events. Keywords: nanotopography, inflammation, osteopontin, SPARC, stabilin 1, cytokines
- Published
- 2018
27. Training Neural Networks from Scratch with Parallel Low-Rank Adapters
- Author
-
Huh, Minyoung, Cheung, Brian, Bernstein, Jeremy, Isola, Phillip, and Agrawal, Pulkit
- Subjects
Computer Science - Machine Learning ,Computer Science - Artificial Intelligence ,Computer Science - Computer Vision and Pattern Recognition - Abstract
The scalability of deep learning models is fundamentally limited by computing resources, memory, and communication. Although methods like low-rank adaptation (LoRA) have reduced the cost of model finetuning, its application in model pre-training remains largely unexplored. This paper explores extending LoRA to model pre-training, identifying the inherent constraints and limitations of standard LoRA in this context. We introduce LoRA-the-Explorer (LTE), a novel bi-level optimization algorithm designed to enable parallel training of multiple low-rank heads across computing nodes, thereby reducing the need for frequent synchronization. Our approach includes extensive experimentation on vision transformers using various vision datasets, demonstrating that LTE is competitive with standard pre-training.
- Published
- 2024
28. A Vision Check-up for Language Models
- Author
-
Sharma, Pratyusha, Shaham, Tamar Rott, Baradad, Manel, Fu, Stephanie, Rodriguez-Munoz, Adrian, Duggal, Shivam, Isola, Phillip, and Torralba, Antonio
- Subjects
Computer Science - Computer Vision and Pattern Recognition ,Computer Science - Computation and Language ,Computer Science - Machine Learning - Abstract
What does learning to model relationships between strings teach large language models (LLMs) about the visual world? We systematically evaluate LLMs' abilities to generate and recognize an assortment of visual concepts of increasing complexity and then demonstrate how a preliminary visual representation learning system can be trained using models of text. As language models lack the ability to consume or output visual information as pixels, we use code to represent images in our study. Although LLM-generated images do not look like natural images, results on image generation and the ability of models to correct these generated images indicate that precise modeling of strings can teach language models about numerous aspects of the visual world. Furthermore, experiments on self-supervised visual representation learning, utilizing images generated with text models, highlight the potential to train vision models capable of making semantic assessments of natural images using just LLMs.
- Published
- 2024
29. Energy Crisis in Nigeria: Evidence from Lagos State
- Author
-
Isola Wakeel Atanda, Mesagan Ekundayo Peter, and Alimi Olorunfemi Yasiru
- Subjects
urbanization ,energy crisis ,electricity ,Nigeria ,Business ,HF5001-6182 ,Economics as a science ,HB71-74 - Abstract
This study examined the nexus between urbanization and energy crisis in Lagos State using questionnaire to elicit information from one hundred randomly selected Lagos residents in Akoka and Onike areas. The study identified the energy supply gap within the Lagos metropolis and found that the energy requirement in the state for industrial, commercial, and domestic needs is not met by supply. It was observed that urbanisation is the root cause of energy crisis in the state and generators provides the main coping strategy for residents. It is therefore recommended that Lagos state should be allowed to generate its own electricity and distribute appropriately among the residents as against the current practice of contributing its generated power to the national grid.
- Published
- 2017
30. Changes in hormonal profiles during competition preparation in physique athletes
- Author
-
Isola, Ville, Hulmi, Juha J., Mbay, Theo, Kyröläinen, Heikki, Häkkinen, Keijo, Ahola, Vilho, Helms, Eric R., and Ahtiainen, Juha P.
- Published
- 2024
- Full Text
- View/download PDF
31. The Role of Medical Thoracoscopy with Talc Poudrage in Spontaneous, Iatrogenic, and Traumatic Pneumothorax: A Prolonged Experience of a Tertiary Care Center
- Author
-
Fantin, Alberto, Castaldo, Nadia, Crisafulli, Ernesto, Sartori, Giulia, Aujayeb, Avinash, Vailati, Paolo, Morana, Giuseppe, Patrucco, Filippo, de Martino, Maria, Isola, Miriam, and Patruno, Vincenzo
- Published
- 2024
- Full Text
- View/download PDF
32. A pilot study evaluating dosing tolerability of 17α-estradiol in male common marmosets (Callithrix jacchus)
- Author
-
Sathiaseelan, Roshini, Isola, Jose V. V., Santín-Márquez, Roberto, Adekunbi, Daniel, Fornalik, Michal, Salmon, Adam B., and Stout, Michael B.
- Published
- 2024
- Full Text
- View/download PDF
33. Transcriptomic analysis of intestine following administration of a transglutaminase 2 inhibitor to prevent gluten-induced intestinal damage in celiac disease
- Author
-
Dotsenko, Valeriia, Tewes, Bernhard, Hils, Martin, Pasternack, Ralf, Isola, Jorma, Taavela, Juha, Popp, Alina, Sarin, Jani, Huhtala, Heini, Hiltunen, Pauliina, Zimmermann, Timo, Mohrbacher, Ralf, Greinwald, Roland, Lundin, Knut E. A., Schuppan, Detlef, Mäki, Markku, and Viiri, Keijo
- Published
- 2024
- Full Text
- View/download PDF
34. Learning Vision from Models Rivals Learning Vision from Data
- Author
-
Tian, Yonglong, Fan, Lijie, Chen, Kaifeng, Katabi, Dina, Krishnan, Dilip, and Isola, Phillip
- Subjects
Computer Science - Computer Vision and Pattern Recognition - Abstract
We introduce SynCLR, a novel approach for learning visual representations exclusively from synthetic images and synthetic captions, without any real data. We synthesize a large dataset of image captions using LLMs, then use an off-the-shelf text-to-image model to generate multiple images corresponding to each synthetic caption. We perform visual representation learning on these synthetic images via contrastive learning, treating images sharing the same caption as positive pairs. The resulting representations transfer well to many downstream tasks, competing favorably with other general-purpose visual representation learners such as CLIP and DINO v2 in image classification tasks. Furthermore, in dense prediction tasks such as semantic segmentation, SynCLR outperforms previous self-supervised methods by a significant margin, e.g., improving over MAE and iBOT by 6.2 and 4.3 mIoU on ADE20k for ViT-B/16., Comment: Code is available at https://github.com/google-research/syn-rep-learn
- Published
- 2023
35. Scaling Laws of Synthetic Images for Model Training ... for Now
- Author
-
Fan, Lijie, Chen, Kaifeng, Krishnan, Dilip, Katabi, Dina, Isola, Phillip, and Tian, Yonglong
- Subjects
Computer Science - Computer Vision and Pattern Recognition - Abstract
Recent significant advances in text-to-image models unlock the possibility of training vision systems using synthetic images, potentially overcoming the difficulty of collecting curated data at scale. It is unclear, however, how these models behave at scale, as more synthetic data is added to the training set. In this paper we study the scaling laws of synthetic images generated by state of the art text-to-image models, for the training of supervised models: image classifiers with label supervision, and CLIP with language supervision. We identify several factors, including text prompts, classifier-free guidance scale, and types of text-to-image models, that significantly affect scaling behavior. After tuning these factors, we observe that synthetic images demonstrate a scaling trend similar to, but slightly less effective than, real images in CLIP training, while they significantly underperform in scaling when training supervised image classifiers. Our analysis indicates that the main reason for this underperformance is the inability of off-the-shelf text-to-image models to generate certain concepts, a limitation that significantly impairs the training of image classifiers. Our findings also suggest that scaling synthetic data can be particularly effective in scenarios such as: (1) when there is a limited supply of real images for a supervised problem (e.g., fewer than 0.5 million images in ImageNet), (2) when the evaluation dataset diverges significantly from the training data, indicating the out-of-distribution scenario, or (3) when synthetic data is used in conjunction with real images, as demonstrated in the training of CLIP models.
- Published
- 2023
36. Neural MMO 2.0: A Massively Multi-task Addition to Massively Multi-agent Learning
- Author
-
Suárez, Joseph, Isola, Phillip, Choe, Kyoung Whan, Bloomin, David, Li, Hao Xiang, Pinnaparaju, Nikhil, Kanna, Nishaanth, Scott, Daniel, Sullivan, Ryan, Shuman, Rose S., de Alcântara, Lucas, Bradley, Herbie, Castricato, Louis, You, Kirsty, Jiang, Yuhao, Li, Qimai, Chen, Jiaxin, and Zhu, Xiaolong
- Subjects
Computer Science - Artificial Intelligence ,Computer Science - Machine Learning ,Computer Science - Multiagent Systems - Abstract
Neural MMO 2.0 is a massively multi-agent environment for reinforcement learning research. The key feature of this new version is a flexible task system that allows users to define a broad range of objectives and reward signals. We challenge researchers to train agents capable of generalizing to tasks, maps, and opponents never seen during training. Neural MMO features procedurally generated maps with 128 agents in the standard setting and support for up to. Version 2.0 is a complete rewrite of its predecessor with three-fold improved performance and compatibility with CleanRL. We release the platform as free and open-source software with comprehensive documentation available at neuralmmo.github.io and an active community Discord. To spark initial research on this new platform, we are concurrently running a competition at NeurIPS 2023.
- Published
- 2023
37. The NeurIPS 2022 Neural MMO Challenge: A Massively Multiagent Competition with Specialization and Trade
- Author
-
Liu, Enhong, Suarez, Joseph, You, Chenhui, Wu, Bo, Chen, Bingcheng, Hu, Jun, Chen, Jiaxin, Zhu, Xiaolong, Zhu, Clare, Togelius, Julian, Mohanty, Sharada, Hong, Weijun, Du, Rui, Zhang, Yibing, Wang, Qinwen, Li, Xinhang, Yuan, Zheng, Li, Xiang, Huang, Yuejia, Zhang, Kun, Yang, Hanhui, Tang, Shiqi, and Isola, Phillip
- Subjects
Computer Science - Artificial Intelligence ,Computer Science - Machine Learning ,Computer Science - Multiagent Systems - Abstract
In this paper, we present the results of the NeurIPS-2022 Neural MMO Challenge, which attracted 500 participants and received over 1,600 submissions. Like the previous IJCAI-2022 Neural MMO Challenge, it involved agents from 16 populations surviving in procedurally generated worlds by collecting resources and defeating opponents. This year's competition runs on the latest v1.6 Neural MMO, which introduces new equipment, combat, trading, and a better scoring system. These elements combine to pose additional robustness and generalization challenges not present in previous competitions. This paper summarizes the design and results of the challenge, explores the potential of this environment as a benchmark for learning methods, and presents some practical reinforcement learning training approaches for complex tasks with sparse rewards. Additionally, we have open-sourced our baselines, including environment wrappers, benchmarks, and visualization tools for future research.
- Published
- 2023
38. LangNav: Language as a Perceptual Representation for Navigation
- Author
-
Pan, Bowen, Panda, Rameswar, Jin, SouYoung, Feris, Rogerio, Oliva, Aude, Isola, Phillip, and Kim, Yoon
- Subjects
Computer Science - Computer Vision and Pattern Recognition ,Computer Science - Artificial Intelligence ,Computer Science - Computation and Language ,Computer Science - Robotics - Abstract
We explore the use of language as a perceptual representation for vision-and-language navigation (VLN), with a focus on low-data settings. Our approach uses off-the-shelf vision systems for image captioning and object detection to convert an agent's egocentric panoramic view at each time step into natural language descriptions. We then finetune a pretrained language model to select an action, based on the current view and the trajectory history, that would best fulfill the navigation instructions. In contrast to the standard setup which adapts a pretrained language model to work directly with continuous visual features from pretrained vision models, our approach instead uses (discrete) language as the perceptual representation. We explore several use cases of our language-based navigation (LangNav) approach on the R2R VLN benchmark: generating synthetic trajectories from a prompted language model (GPT-4) with which to finetune a smaller language model; domain transfer where we transfer a policy learned on one simulated environment (ALFRED) to another (more realistic) environment (R2R); and combining both vision- and language-based representations for VLN. Our approach is found to improve upon baselines that rely on visual features in settings where only a few expert trajectories (10-100) are available, demonstrating the potential of language as a perceptual representation for navigation.
- Published
- 2023
39. Class-wide genomic tendency throughout specific extremes in black fungi
- Author
-
Coleine, Claudia, Kurbessoian, Tania, Calia, Giulia, Delgado-Baquerizo, Manuel, Cestaro, Alessandro, Pindo, Massimo, Armanini, Federica, Asnicar, Francesco, Isola, Daniela, Segata, Nicola, Donati, Claudio, Stajich, Jason E, de Hoog, Sybren, and Selbmann, Laura
- Subjects
Biological Sciences ,Genetics ,Biotechnology ,Human Genome ,Health Disparities ,Black fungi ,Stress resistance ,Comparative genomics ,Extreme environments ,Evolutionary Biology ,Microbiology ,Plant Biology ,Mycology & Parasitology ,Evolutionary biology ,Plant biology - Published
- 2024
40. Food safety knowledge and attitudes among fish vendors in informal markets in Ilorin, Nigeria: a cross-sectional study
- Author
-
Ghali-Mohammed, Ibraheem, Isola, Tajudeen Opeyemi, Adeyemo, Ismail Adewuyi, Kadir, Rafiu Adebisi, Ambali, Hawau Motunrayo, Alhaji, Nma Bida, and Odetokun, Ismail Ayoade
- Published
- 2024
- Full Text
- View/download PDF
41. Effect of different anticoagulants and antiplatelets on intraoral bleeding time during professional oral hygiene session
- Author
-
Pesce, Paolo, Pin, Ludovica, Pin, Daniele, Bagnasco, Francesco, Ball, Lorenzo, Isola, Gaetano, Nicolini, Paolo, and Menini, Maria
- Published
- 2024
- Full Text
- View/download PDF
42. The role of cellular senescence in ovarian aging
- Author
-
Hense, Jéssica D., Isola, José V. V., Garcia, Driele N., Magalhães, Larissa S., Masternak, Michal M., Stout, Michael B., and Schneider, Augusto
- Published
- 2024
- Full Text
- View/download PDF
43. Proteins from shrews’ venom glands play a role in gland functioning and venom production
- Author
-
Kowalski, Krzysztof, Marciniak, Paweł, Nekaris, K. Anne-Isola, and Rychlik, Leszek
- Published
- 2024
- Full Text
- View/download PDF
44. Fifteen-year recall period on zirconia-based single crowns and fixed dental prostheses. A prospective observational study
- Author
-
Khijmatgar, Shahnawaz, Tumedei, Margherita, Tartaglia, Guilia, Crescentini, Michele, Isola, Gaetano, Sidoti, Ernesto, Sforza, Chiarella, Del Fabbro, Massimo, and Tartaglia, Gianluca Martino
- Published
- 2024
- Full Text
- View/download PDF
45. Enteric Glia Mediate Neuron Death in Colitis Through Purinergic Pathways That Require Connexin-43 and Nitric OxideSummary
- Author
-
Isola A.M. Brown, Jonathon L. McClain, Ralph E. Watson, Bhavik A. Patel, and Brian D. Gulbransen
- Subjects
Diseases of the digestive system. Gastroenterology ,RC799-869 - Abstract
Background & Aims: The concept of enteric glia as regulators of intestinal homeostasis is slowly gaining acceptance as a central concept in neurogastroenterology. Yet how glia contribute to intestinal disease is still poorly understood. Purines generated during inflammation drive enteric neuron death by activating neuronal P2X7 purine receptors (P2X7R); triggering adenosine triphosphate (ATP) release via neuronal pannexin-1 channels that subsequently recruits intracellular calcium ([Ca2+]i) in surrounding enteric glia. We tested the hypothesis that the activation of enteric glia contributes to neuron death during inflammation. Methods: We studied neuroinflammation in vivo using the 2,4-dinitrobenzene sulfonic acid model of colitis and in situ using whole-mount preparations of human and mouse intestine. Transgenic mice with a targeted deletion of glial connexin-43 (Cx43) [GFAP::CreERT2+/â/Cx43f/f] were used to specifically disrupt glial signaling pathways. Mice deficient in inducible nitric oxide (NO) synthase (iNOSâ/â) were used to study NO production. Protein expression and oxidative stress were measured using immunohistochemistry and in situ Ca2+ and NO imaging were used to monitor glial [Ca2+]i and [NO]i. Results: Purinergic activation of enteric glia drove [Ca2+]i responses and enteric neuron death through a Cx43-dependent mechanism. Neurotoxic Cx43 activity, driven by NO production from glial iNOS, was required for neuron death. Glial Cx43 opening liberated ATP and Cx43-dependent ATP release was potentiated by NO. Conclusions: Our results show that the activation of glial cells in the context of neuroinflammation kills enteric neurons. Mediators of inflammation that include ATP and NO activate neurotoxic pathways that converge on glial Cx43 hemichannels. The glial response to inflammatory mediators might contribute to the development of motility disorders. Keywords: Enteric Nervous System, Hemichannels, Oxidative Stress, Purines
- Published
- 2016
- Full Text
- View/download PDF
46. Benchmarking Robustness and Generalization in Multi-Agent Systems: A Case Study on Neural MMO
- Author
-
Chen, Yangkun, Suarez, Joseph, Zhang, Junjie, Yu, Chenghui, Wu, Bo, Chen, Hanmo, Zhu, Hengman, Du, Rui, Qian, Shanliang, Liu, Shuai, Hong, Weijun, He, Jinke, Zhang, Yibing, Zhao, Liang, Zhu, Clare, Togelius, Julian, Mohanty, Sharada, Chen, Jiaxin, Li, Xiu, Zhu, Xiaolong, and Isola, Phillip
- Subjects
Computer Science - Artificial Intelligence - Abstract
We present the results of the second Neural MMO challenge, hosted at IJCAI 2022, which received 1600+ submissions. This competition targets robustness and generalization in multi-agent systems: participants train teams of agents to complete a multi-task objective against opponents not seen during training. The competition combines relatively complex environment design with large numbers of agents in the environment. The top submissions demonstrate strong success on this task using mostly standard reinforcement learning (RL) methods combined with domain-specific engineering. We summarize the competition design and results and suggest that, as an academic community, competitions may be a powerful approach to solving hard problems and establishing a solid benchmark for algorithms. We will open-source our benchmark including the environment wrapper, baselines, a visualization tool, and selected policies for further research.
- Published
- 2023
47. Distilled Feature Fields Enable Few-Shot Language-Guided Manipulation
- Author
-
Shen, William, Yang, Ge, Yu, Alan, Wong, Jansen, Kaelbling, Leslie Pack, and Isola, Phillip
- Subjects
Computer Science - Computer Vision and Pattern Recognition ,Computer Science - Artificial Intelligence ,Computer Science - Computation and Language ,Computer Science - Machine Learning ,Computer Science - Robotics - Abstract
Self-supervised and language-supervised image models contain rich knowledge of the world that is important for generalization. Many robotic tasks, however, require a detailed understanding of 3D geometry, which is often lacking in 2D image features. This work bridges this 2D-to-3D gap for robotic manipulation by leveraging distilled feature fields to combine accurate 3D geometry with rich semantics from 2D foundation models. We present a few-shot learning method for 6-DOF grasping and placing that harnesses these strong spatial and semantic priors to achieve in-the-wild generalization to unseen objects. Using features distilled from a vision-language model, CLIP, we present a way to designate novel objects for manipulation via free-text natural language, and demonstrate its ability to generalize to unseen expressions and novel categories of objects., Comment: Project website at https://f3rm.csail.mit.edu, Accepted at the 7th Annual Conference on Robot Learning (CoRL), 2023 in Atlanta, US
- Published
- 2023
48. The “Relevance” of the African Traditional Medicine (Alternative Medicine) to Health Care Delivery System in Nigeria
- Author
-
Isola, Omoleke Ishaq
- Published
- 2013
- Full Text
- View/download PDF
49. Senolytic treatment fails to improve ovarian reserve or fertility in female mice
- Author
-
Garcia, Driele N., Hense, Jessica D., Zanini, Bianka M., Isola, Jose V. V., Prosczek, Juliane B., Ashiqueali, Sarah, Oliveira, Thais L., Mason, Jeffrey B., Schadock, Ines C., Barros, Carlos C., Stout, Michael B., Masternak, Michal M., and Schneider, Augusto
- Published
- 2024
- Full Text
- View/download PDF
50. DreamSim: Learning New Dimensions of Human Visual Similarity using Synthetic Data
- Author
-
Fu, Stephanie, Tamir, Netanel, Sundaram, Shobhita, Chai, Lucy, Zhang, Richard, Dekel, Tali, and Isola, Phillip
- Subjects
Computer Science - Computer Vision and Pattern Recognition ,Computer Science - Machine Learning - Abstract
Current perceptual similarity metrics operate at the level of pixels and patches. These metrics compare images in terms of their low-level colors and textures, but fail to capture mid-level similarities and differences in image layout, object pose, and semantic content. In this paper, we develop a perceptual metric that assesses images holistically. Our first step is to collect a new dataset of human similarity judgments over image pairs that are alike in diverse ways. Critical to this dataset is that judgments are nearly automatic and shared by all observers. To achieve this we use recent text-to-image models to create synthetic pairs that are perturbed along various dimensions. We observe that popular perceptual metrics fall short of explaining our new data, and we introduce a new metric, DreamSim, tuned to better align with human perception. We analyze how our metric is affected by different visual attributes, and find that it focuses heavily on foreground objects and semantic content while also being sensitive to color and layout. Notably, despite being trained on synthetic data, our metric generalizes to real images, giving strong results on retrieval and reconstruction tasks. Furthermore, our metric outperforms both prior learned metrics and recent large vision models on these tasks., Comment: Website: https://dreamsim-nights.github.io/ Code: https://github.com/ssundaram21/dreamsim
- Published
- 2023
Catalog
Discovery Service for Jio Institute Digital Library
For full access to our library's resources, please sign in.