dataset

TPG
2023-11
Researchers from the University of Chicago introduce Nightshade, an optimized prompt-specific poisoning attack where poison samples look visually identical to benign images with matching text prompts.
TPG
2023-11
A new research paper that produces consistent multi-view normal maps and correpsonding color images, and thus reconstructs high-fidelity textured mesh from a single image in only 2~3 minutes.
Nikki Stevens and and OS Keyes situate FRT as an infrastructural assemblage and focus on the history of four facial recognition datasets
Sofia Crespo and Anna Ridler unpick technologies and how they are created, and this process of building the dataset starts to think through some of the implications inherent in even analogue …
What Does the Algorithm See? Panellists, artist / theorist / curator Rosa Menkman and artist Joanna Zylinska join Dr Rachel O’Dwyer, NCAD.
The company's star ethics researcher highlighted the risks of large language models, which are key to Google's business.
TPG
2020-10
“Question to twitterverse: A lovely PhD student and I are looking for papers and other projects from the social sciences and humanities on computer vision, image recognition and on the production of …
We propose Localized Narratives, a new form of multimodal image annotations connecting vision and language. We ask annotators to describe an image with their voice while simultaneously hovering their …
June 29th, 2020 It has been brought to our attention [1] that the Tiny Images dataset contains some derogatory terms as categories and offensive images. This was a consequence of the automated data …
We propose Localized Narratives, an efficient way to collect image captions with dense visual grounding. We ask annotators to describe an image with their voice while simultaneously hovering their …
However, thanks to the pandemic, a few British hospitals are now rolling out AI tools to help medical staff interpret chest X-rays more quickly. For instance, staff at the Royal Bolton Hospital, are …
The face of your voice 3D, from the verbal to the physiognomic Contemporary life seems to be an endless game of data quantification, moving across different cultural domains. The former is an …
The Earth Archive The Earth Archive is both a program of scanning focused on endangered landscapes and an open-source collection of LiDAR scans accessible to scientists around the world. Our …
the critical dictionary of southeast asia (cdosea), begins with a question: what constitutes the unity of southeast asia — a region never unified by language, religion or political power?cdosea …
TPG
2020-04
Calm Technology IMPAKT Festival 2019 Panel discussion: Calm Technology Speakers: Olia Lialina, David Benqué & Cristina Cochior. Moderator: Annet Dekker In the 90s the concept of calm technology …
Datasets are large collections of digital information that are used to train AI. They might contain anything from weather data, such as air pressure and temperature, to photos, music, or indeed …
A popular self-driving car dataset is missing labels for hundreds of pedestrians https://blog.roboflow.ai/self-driving-car-dataset-missing-pedestrians/
The Next Biennial Should be Curated by a Machine is an inquiry into the relationship between curating and artificial intelligence, and a possibility of developing an experimental system capable of …
Early Modern Computer Vision - Leonardo Impett https://docs.google.com/document/d/1LKs82uKkSgQ-4wGUQ4Dwzxgnerx2e6zbHf4iGIHuJmI/edit#heading=h.60chgdizcy6h
Egor Tsvetkov - Your Face Is Big Data The next time you ride the subway in St. Petersburg, watch out for 21-year-old photographer Egor Tsvetkov. He recently unveiled a new project called “YOUR FACE …
Reconstructing 3D human shape and pose from a monocular image Reconstructing 3D human shape and pose from a monocular image is challenging despite the promising results achieved by the most recent …
Until recently, Hoan Ton-That’s greatest hits included an obscure iPhone game and an app that let people put Donald Trump’s distinctive yellow hair on their own photos. Then Mr. Ton-That — an …
In Heather Dewey-Hagborg’s artwork ‘How do you see me?’, commissioned for the Data/Set/Match programme at The Photographers’ Gallery, the artist explores how machines see us. A question that has been carefully slipping through several areas of production and research during the past couple of decades. At the same time an essential need has also …
An introductory presentation about Data / Set / Match, a year-long programme seeking new ways to present, visualise and interrogate contemporary image datasets. Departing from traditional 19th and …
Dare to share your poop for science—and help change the future of gut health. We’re building the world’s first and largest poop image database–so we can train an AI to change the future …
Using generative adversarial networks (GAN), we can learn how to create realistic-looking fake versions of almost anything, as shown by this collection of sites.
TPG
2019-10
Ramon Amaro introduces the basics of machine learning, its criteria for assigning value, the collision between blackness and the artificial, its flaws, and the problem of impunity that all too often …
TPG
2019-09
Generated Photos 100,000 Faces Generated by AI Free to Download These people aren’t real! We are building the next generation of media through the power of AI (an original machine learning …
What do you see, YOLO9000? by Taller Estampa | Soy Cámara YOLO9000 is a trained object recognition neuronal network with a dataset of 9,418 words and millions of images. It is one of the many …
To mitigate the race bias in these datasets, we construct a novel face image dataset, containing 108,501 images, with an emphasis of balanced race composition in the dataset. We define 7 race groups: …
We introduce the first visual privacy dataset originating from people who are blind in order to better understand their privacy disclosures and to encourage the development of algorithms that can …
Visual Dialog requires an AI agent to hold a meaningful dialogue with humans in natural, conversational language about visual content. Specifically, given an image, a dialogue history, and a …
We introduce natural adversarial examples – real-world, unmodified, and naturally occurring examples that cause classifier accuracy to significantly degrade. We curate 7,500 natural adversarial …
As researchers, we have always wondered: if we scale up the amount of training data 10x, will the accuracy double? https://ai.googleblog.com/2017/07/revisiting-unreasonable-effectiveness.html …
High Quality Face Recognition with Deep Metric Learning The new example comes with pictures of bald Hollywood action heroes and uses the provided deep metric model to identify how many different …
The reason biases against women or people of colour appear in technology are complex. They’re often attributed to data sets being incomplete and the fact that the technology is often made by people …
This labelling job has made me very observant. I have found pictures that made me think “if I had taken such a picture, then I would know what is everything.”  For instance, in a picture of a …
We knew we needed to collect a data set that has far more images than we have ever had before, perhaps thousands of times more, and together with Professor Kai Li at Princeton University, we launched …
Let’s play  Name That Dataset!!! https://people.csail.mit.edu/torralba/research/bias/
i will tell you everything Manetta Berends, 2015 training set = “contemporary encyclopaedia” In the process of making an encyclopaedia, categories are decided on wherein various objects …
VFRAME Adam Harvey A collection of open-source computer vision software tools designed specifically for human rights investigations that rely on large datasets of visual media. Specifically VFRAME is …
ImageNet Roulette (Trevor Paglen, 2019) uses a neural network trained on the “people” categories from the ImageNet dataset to classify pictures of people. It’s meant to be a peek into how artificial …
What can algorithms know? The present power of algorithms is fueled by another entity: that of data. Generally referred to as big data, large data sets, whose technical history has been well …
For the past six years Heather Dewey-Hagborg has been researching, writing and producing artwork engaging the methodology of ‘forensic DNA phenotyping’. In this essay, she explores a different aspect of this technology and questions: is forensic DNA phenotyping a photographic process?
“AI, Ain’t I A Woman ” - a spoken word piece that highlights the ways in which artificial intelligence can misinterpret the images of iconic black women: Oprah, Serena Williams, …
Explore Airbnb listings through algorithmically generated collages.
Abnormality Detection in Images When describing images, humans tend not to talk about the obvious, but rather mention what they find interesting. We argue that abnormalities and deviations from …