
Sade Olutola
𓃗

PR's Tumblrdome
cherry valley forever
EXPECTATIONS
Jules of Nature

Andulka
$LAYYYTER
KIROKAZE
Misplaced Lens Cap
sheepfilms
Game of Thrones Daily
official daine visual archive

gracie abrams

Product Placement
let's talk about Bridgerton tea, my ask is open
will byers stan first human second
almost home

seen from United Kingdom

seen from United Kingdom

seen from Argentina

seen from United States

seen from Malaysia
seen from United States

seen from Malaysia

seen from Netherlands
seen from South Africa

seen from United States
seen from Malaysia

seen from Germany

seen from United States
seen from Venezuela

seen from Saudi Arabia
seen from United States
seen from United States
seen from Brazil

seen from Germany

seen from Canada
@xinran-leiyu

Anya is live and ready to show you everything. Watch her strip, dance, and perform exclusive shows just for you. Interact in real-time and make your fantasies come true.
Free to watch • No registration required • HD streaming
Use novel content to subtly imply during visual data extraction (All 5 Sites)

Yarbus’s classic eye-tracking experiment demonstrated that different viewing tasks and question framings direct observers’ attention to different regions of the same image, resulting in markedly different visual scanpaths.



Matched paintings and painting fragments
The keywords of the novel text match the fragments of the painting.
The generated 3D model
Sculpting by Gaze: A Narrative in Every Window
Match results
Model generation
How does the dataset influence architectural form?

Anya is live and ready to show you everything. Watch her strip, dance, and perform exclusive shows just for you. Interact in real-time and make your fantasies come true.
Free to watch • No registration required • HD streaming
The entire venue
Virtual reality viewing trajectory of the venue
Site thermal vision analysis
matched the datasets
Design Site selection
Matching logic
This paper discusses visual saliency analysis - we trained the SOM based on visual attractiveness features.
som(film and airbnb)
This example links the real photos of Airbnb from the two pictures with the dataset of movie scenes.
Why use "som"? After som clustering, a cell contains several similar data. In the next part, this logic is used to find the most similar cell for matching. Essentially, it is based on the multi-match feature after machine learning clustering.
Four databases
Artwork data:
1,000 paintings of windows were crawled from multiple websites such as CMA, Wikiart, and Met.
Film
collected from Panoramas of Cinema
Novel
From 3544 Dutch novels, 1000 story segments strongly related to windows
from two website:
gutenberg project
National Library of the Netherlands (KB).
Airbnb
Windows Watching
Theoretical Framework
Preparations
1.Site Model Processing ( Finish )
2.Site Matching ( Finish )
3.Transforming painting fragments into 3D (subsequently creating abstract forms using depth maps)
Later, we plan to use the matched movie textures and furniture extracted from depth-influenced paintings, then insert them into the room.

Anya is live and ready to show you everything. Watch her strip, dance, and perform exclusive shows just for you. Interact in real-time and make your fantasies come true.
Free to watch • No registration required • HD streaming
Text → Data → Space
Gaze interference model
Site Selection
The project outcomes can take two directions.
One approach involves reimagining the rooms behind facades through a voyeuristic lens and presenting them to the public: conducting VR line-of-sight scans of Rotterdam’s actual building facades, recording how long contemporary passersby gaze at each window. Windows where gaze duration exceeds a certain threshold are deemed "activated." Each activated window is then take over by a cluster of Airbnb data corresponding to its geographic coordinates—through cross-modal retrieval, a set of culturally bound signatures (movie frames, painting color palettes, novel excerpts) is assigned to it. The original room behind the window is reimagined: using these signatures, a new small room is generated as a secondary narrative of the story hidden behind that window. These newly infused rooms leave visible traces on the façade detectable by pedestrians, inviting viewers from the street to peer into a museum scattered across the city's real-world buildings.
The second is extending the original building's facade: scaling up the parasitic dimension to form an entire layer of public space at the city's top. In the most densely activated neighborhoods, each building’s roof is governed by a set of corresponding Airbnb data—adding an imaginary "window narrative layer." The activated windows are here Boolean-cut and lifted as a whole to the rooftop, becoming entrances and exterior elements of this new level. Each small room on the roof transforms the previously voyeuristic privacy into a shared public space where viewing becomes mutual—where those who look are also seen looking. Thus, private windows grow into the city’s public uppermost layer.
Windows: The Urban Interface of Seeing and Being Seen
A window is the oldest public-private apparatus of the city. It dictates who may look in, who may look out, what is put on display, and what is concealed—an entire urban order built around intimacy, etiquette, surveillance and theatrical performance has settled upon this thin sheet of glass over centuries. Dutch urban culture has sharpened this apparatus to an exceptional edge: from the unobscured wide windows of Calvinist "transparency without concealment", to interiors sliced by window light in the canvases of Vermeer and de Hooch, and onward to the red-light district windows rented out for paid spectatorship. The window bears enduring traces of Rotterdam’s perpetual negotiation over the dynamic of seeing and being seen.
Airbnb has torn away this delicate negotiation. Once a room is photographed and uploaded online, the entire interior transforms into an outward-facing screen. The regulation of visibility shifts from architectural elements—curtains, room depth, thresholds, natural light—to image production: framing, artificial lighting, and the choice of whether to post the space online at all. Private domestic spaces are converted into globally viewable paid imagery, upending the boundary between public and private: what lies inside becomes far more exposed than what lies outside.
This project acknowledges this new reality yet rejects its passive surrender. It does not treat the window as a relic to be nostalgically restored to its former function, but frames it as the city’s last remaining public-private interface still open to design, habitation and renegotiation—a space where narratives of the windowsill may unfold once more, rather than being bypassed entirely. Drawing on a dataset of 1,084 Airbnb interiors across Rotterdam, the work extracts the gazes, narratives, colour palettes and scenes inherent to every window from cultural moulds of urban storytelling: film, literature, painting and photography. Deep learning algorithms then quantify the intensity of the stares each of these windows once endured. Windows stand as the city’s last original public-private apparatus. Rather than mending their vanished past, this project harnesses the stories behind one thousand and eighty-four windows to redesign their future.

Anya is live and ready to show you everything. Watch her strip, dance, and perform exclusive shows just for you. Interact in real-time and make your fantasies come true.
Free to watch • No registration required • HD streaming
Matching Guide - Three Colors Correspond to Three Datasets
We first mapped out the entire matching process, refined our visualization approach, and clarified our design philosophy—returning privacy to the public through the stories behind a thousand windows.
Painting Matching - Spatial Color
Last week's painting matching logic was unclear because the artwork descriptions were projected onto the novel SOM and then matched with Airbnb descriptions, involving multiple intermediate conversions, resulting in weak alignment between the matched paintings and Airbnb space results. A simpler approach using basic colors and color ratios for matching would be better.