简体中文
EnglishWebsite review
This review may include AI-generated content and is not individually verified. Domain, certificate and other facts come from the observed data on this page.
What is Computer Vision Art Gallery?
Computer Vision Art Gallery is an online exhibition space for artwork made with or about computer vision. It is tied to the CVPR conference; the 2021 edition was virtual because of COVID-19, and the site also keeps past galleries.
Its structure is simple: a winner, a shortlist, and a broader list of accepted works. That makes it useful less as a browsable museum and more as a curated snapshot of how artists, researchers and creative technologists were using machine vision at a particular moment.
What you will find
- A single winning work, highlighted at the top.
- A shortlist of notable pieces, with artist names, titles and years.
- A longer list of accepted works, again with artist, title and year.
- Past gallery sections, so you can compare editions over time.
The works mix technical and conceptual approaches. Some, like Alexander Mordvintsev's Hexells, come from people known for computer-vision research. Others, such as Jake Elwes's The Zizi Show or Sofia Crespo and Entangled Others' Beneath the Neural Waves, sit closer to media art and generative art. Titles like Faces2voices, Fear Filter and Paint Your Face Away suggest recurring themes: surveillance, identity, facial recognition and what happens when machines interpret human images.
Who it is for
- Artists looking for precedents and peer work in CV-based art.
- Researchers and students who want a cultural frame for vision systems.
- Curators or event organisers seeking names and project examples.
- Anyone curious about AI art beyond generic image generators.
A practical next step: open the shortlist first, pick two or three titles that interest you, and note the year and artist. Then search those artists' own sites to see fuller documentation, since this gallery mainly presents titles and names rather than deep artist statements. If you are submitting or curating, use the past galleries to judge the range of accepted work and the balance between technical demos and critical pieces.
For context on the conference itself, see CVPR.
How can I submit my artwork to the Computer Vision Art Gallery?
The site does not publish a submission process on the page you're asking about. What it shows is a gallery program tied to CVPR 2021 (held virtually due to COVID-19), with work organized into a Winner, a Shortlist, and an "All Accepted Works" section, plus a "Past Galleries" area. That structure tells you submissions are juried and edition-based, not an open, always-on upload.
Computer Vision Art Gallery
What the page implies about submitting
- Entry is tied to a specific call and year, so timing matters more than a permanent form.
- Selection is competitive: the page separates a single winner, a shortlist, and a broader accepted set.
- The work shown is technically engaged — titles reference GANs, latent space, surveillance, neural waves, camouflage, and generative systems — so a submission needs a clear concept, not just an output image.
- Accepted pieces appear as titled works with artist names and years, so you'll need a finished title and date.
A practical next step
Look for the current call for entries or contact route on the gallery's own site first, then check the CVPR conference site for the art gallery track, since this gallery is presented alongside CVPR. If no open call is listed, the realistic path is to prepare a submission package and wait for the next edition: a short statement, 3–6 documentation images or a video link, a technical description of the method, and a one-line bio.
Decision criteria before you send anything
| Question | Why it matters |
|---|---|
| Does your piece use computer vision or machine learning as part of the work, not just as a filter? | The gallery's identity is technical practice, not general digital art. |
| Can you explain the method in plain language? | Reviewers need to place the work among generative, surveillance, and neural-network themes. |
| Is the work finished and titled? | Accepted listings show a title, artist, and year. |
| Can you meet a fixed deadline? | Editions are annual and tied to a conference date. |
If your practice fits generative or vision-based systems, this is a reasonable target; if it doesn't, a broader new-media venue will likely be a better match.
What are the criteria for judging artworks in the Computer Vision Art Gallery?
The page does not publish a formal judging rubric. What it does show is a juried exhibition structure: a Winner, a Shortlist, and a larger All Accepted Works list, with each entry credited to an artist and dated. That structure implies a two-stage process — works are first accepted into the gallery, then a smaller set is shortlisted, and one is named winner — but the specific criteria used at each stage aren't stated on the page.
What can be inferred from the accepted works
Looking at the titles and artists, the gallery appears to favour work where computer vision is the material or subject, not just a production tool. Recurring themes include:
- Surveillance and machine looking — David Rokeby's Dundas Square Surveillance Etude, Rod Dickinson's Fear Filter, Shinji Toya's Paint Your Face Away.
- Dataset and classification critique — Alexia Achilleos's Cyprus as AI Saw it in 1879, Matthew Plummer-Fernandez's Cave Paintings.
- Generative and neural-network aesthetics — Sofia Crespo x Entangled Others' Beneath the Neural Waves, Alexander Mordvintsev's Hexells, Daniel Ambrosi's Infinite Dreams.
- Bodies, faces and identity — Jake Elwes's The Zizi Show, Helena Nikonole & Nikita Prudnikov's Faces2voices, yuma kishi's Mary.
A practical reading: a work is more likely to fit if it interrogates how machines see, classify or generate, rather than using CV output as decoration.
How to judge for yourself
If you're deciding whether a piece belongs in this kind of show, ask:
- Is the vision system essential? Would the work collapse if the model were removed?
- Does it say something about the technology? Critique, reframing or exposure of a dataset tends to read as stronger than novelty output.
- Is the concept legible without a caption? Juried shows often favour works that survive a quick pass.
- Does it sit in conversation with the field? The list shows clear lineages — surveillance art, latent-space work, GAN portraiture.
For the actual selection rules, the most reliable next step is to check the gallery's own call for entries or the CVPR workshop page for that year, since criteria typically change per edition. The gallery itself is at Computer Vision Art Gallery.
Can I visit the Computer Vision Art Gallery in person, or is it only virtual?
The Computer Vision Art Gallery is not a physical venue you can visit. It is an online exhibition tied to the CVPR (Computer Vision and Pattern Recognition) conference, and for 2021 the gallery is explicitly virtual due to COVID-19. The page lists past galleries, a winner, a shortlist and all accepted works, which is the typical structure of a curated online showcase rather than a museum with opening hours.
If you want to "visit," treat it as a browsing session: open the gallery, start with the winner and shortlist to see the strongest examples, then move through the accepted works. A useful decision criterion is whether you care about the art itself or the technical methods behind it—the winner and shortlist entries are the quickest way to see what the curators considered most successful, while the full accepted list is better for scanning themes such as surveillance, identity, ecology and machine imagination.
For context on the kind of event it belongs to, see CVPR.
Who won the Computer Vision Art Gallery award in 2021?
Nouf Aljowaysir won with Salaf (2020), listed as the Winner on the Computer Vision Art Gallery's CVPR 2021 page: Computer Vision Art Gallery.
If you want to judge whether the award points to work you'd actually enjoy, compare it with the shortlist rather than treating the win as a single must-see. The 2021 shortlist included Yuguang Zhang's (Non-)Human, Vadim Epstein's Aphantasia, Sofia Crespo x Entangled Others' Beneath the Neural Waves, David Rokeby's Dundas Square Surveillance Etude, and Jake Elwes's The Zizi Show — a useful spread across identity, machine perception, ecology and performance.
What types of computer vision techniques are commonly featured in the gallery's artworks?
The gallery does not present a fixed curriculum of techniques; its accepted works cluster around a few recurring computer-vision and machine-learning methods, usually chosen for what they let an artist say rather than for benchmark performance.
Common technique families in the listed works
- Generative adversarial networks and latent-space exploration. Many titles point to GANs or latent manipulation, e.g. Tyler Liu's IDEOLOGICAL WAR OF GAN, Vít Růžička's Latent Rider, and the group work From Noise to Form.
- Neural style transfer and neural synthesis. Works such as Sofia Crespo x Entangled Others' Beneath the Neural Waves and Alexander Mordvintsev's Hexells sit in the tradition of learned visual synthesis, including cellular-automata-like neural systems.
- Classification, recognition and captioning. Jono Brandel's Des Objets Reconnus ("recognized objects") and Bernat Cuní's C is for CAT engage object recognition and its labels directly.
- Face, gaze and pose analysis. Faces2voices by Helena Nikonole & Nikita Prudnikov, Shinji Toya's Paint Your Face Away, and Hye Min Cho's Fullyconnected-autorefractor use facial or biometric processing.
- Surveillance and tracking. David Rokeby's Dundas Square Surveillance Etude and Rod Dickinson's Fear Filter treat detection and monitoring as subject matter.
- Reconstruction, mapping and 3D/spatial vision. François Quévillon's Mouvant and Ben Bogart's Imagined Field from the Decomposition of an Apparatus work with reconstruction and decomposition of scenes.
- Dataset and archive critique. Alexia Achilleos' Cyprus as AI Saw it in 1879 and Derek Curry & Jennifer Gradecki's Infodemic examine how training data and models shape historical or informational truth.
- Human–AI collaboration and performance. Jake Elwes' The Zizi Show and Meltem Şahin's sketches with+for+from ai foreground the interaction between performer, artist and model.
What this means in practice
If you are an artist deciding what to learn, the gallery's pattern suggests starting with a model you can inspect and iterate on — a classifier or a GAN whose failures are visible — rather than a black-box API. If you are a curator or researcher, the list is more useful as a map of themes (surveillance, identity, ecology, archives) than as a technical taxonomy, since several works combine two or more of the above.
Next step
Open the gallery's "All Accepted Works" list at Computer Vision Art Gallery and sort the pieces by technique you already know; the shortlist and winner entries are a faster way to see which approaches the jury rewarded.
Akamai Connected Cloud
Toronto, Ontario, Canada
User reviews (0)