Text-to-image models (T2I) such as StableDiffusion have been used to generate high quality images of people. However, due the random nature generation process, person has a different appearance e.g. pose, face, and clothing, despite using same text prompt. The inconsistency makes T2I unsuitable for pose transfer. We address this by proposing multimodal diffusion model that accepts text, visual prompting. Our is first unified method perform all image tasks-generation, transfer, mask-less...

We introduce the first generative model capable of simultaneous multi-object compositing, guided by both text and layout. Our allows for addition multiple objects within a scene, capturing range interactions from simple positional relations (e.g., next to, in front of) to complex actions requiring reposing hugging, playing guitar). When an interaction implies additional props, like `taking selfie', our autonomously generates these supporting objects. By jointly training compositing...

We present EKILA; a decentralized framework that enables creatives to receive recognition and reward for their contributions generative AI (GenAI). EKILA proposes robust visual attribution technique combines this with an emerging content provenance standard (C2PA) address the problem of synthetic image – determining model training data responsible AI-generated image. Furthermore, extends non-fungible token (NFT) ecosystem introduce tokenized representation rights, enabling triangular...

The Dayton Agreement continues to be mentioned as a potential model by Western politicians and pundits for various conflicts sectarian violence around the globe, most recently Syria, Ukraine Iraq. Indeed, is rightly associated Bosnians foreigners alike with end of war absence renewed armed conflict. However, people in BiH it also ushering political-economic order inequality dispossession, not only means dignified livelihood, but future agentive capacities shape that future. For this reason,...

ABSTRACT In 2002, Bosnia and Herzegovina was the site of massive housing‐reconstruction projects run by international aid organizations as part a plan to move refugees back their prewar homes. Alongside usual technical tasks such projects, workers spent considerable time effort establishing stabilizing categorical distinction between things humanitarian political. Analysis these efforts one organization reveals often‐overlooked social cultural work necessary maintain field operation,...

In early February 2014, mass unrest shook the streets of Tuzla, a northern city in Bosnia–Herzegovina and eventually forced resignation regional government. Conspicuous at protests were repertoire practices media taken directly from football stadium, which indexed presence participation Ultra fans city’s team. This article contextualizes decisive role played by Tuzla uprising as part historically specific gendered claim to social incorporation right an era youth unemployment postsocialist...

Abstract This article explores the ambivalent forms of authority and legitimacy articulated by Office High Representative international community in postwar Bosnia Herzegovina. The exercised quasi-sovereign powers that placed his position at center two contradictions: a democratization paradox “imposing democracy,” is, promoting democracy through undemocratic means, state-building building an independent state violating principle popular sovereignty. I analyze Office's use mass-mediated...

A disproportionate focus on nostalgia in post-socialist societies has resulted analytic limits to our ability account for diverse uses of the past. I argue that concepts chronotope, voice, and participation framework, when combined with scholarship's attention role emotion affect semiosis, can move us beyond these limits. The resulting framework will be value anthropologists interested relationship between historical consciousness, social action, cultural production.

Neural Style Transfer (NST) is the field of study applying neural techniques to modify artistic appearance a content image match style reference image. Traditionally, NST methods have focused on texture-based edits, affecting mostly low level information and keeping most structures same. However, style-based deformation desirable for some styles, especially in cases where abstract or primary concept its deformed rendition content. With recent introduction diffusion models, such as Stable...

Article Legitimacy Matters: Managing the Democratization Paradox of Foreign State-Building in Bosnia-Herzegovina was published on April 1, 2012 journal Comparative Southeast European Studies (volume 60, issue 4).

Compositing an object into image involves multiple non-trivial sub-tasks such as placement and scaling, color/lighting harmonization, viewpoint/geometry adjustment, shadow/reflection generation. Recent generative compositing methods leverage diffusion models to handle at once. However, existing face limitations due their reliance on masking the original during training, which constrains generation input mask. Furthermore, obtaining accurate mask specifying location scale of in a new can be...

Representation learning aims to discover individual salient features of a domain in compact and descriptive form that strongly identifies the unique characteristics given sample respective its domain. Existing works visual style representation literature have tried disentangle from content during training explicitly. A complete separation between these has yet be fully achieved. Our paper learn artistic more disentangled semantic depicted an image. We use Neural Style Transfer (NST) measure...

Text-to-image models (T2I) such as StableDiffusion have been used to generate high quality images of people. However, due the random nature generation process, person has a different appearance e.g. pose, face, and clothing, despite using same text prompt. The inconsistency makes T2I unsuitable for pose transfer. We address this by proposing multimodal diffusion model that accepts text, visual prompting. Our is first unified method perform all image tasks - generation, transfer, mask-less...