Photon
photonsearch.azurewebsites.net
- Windows
- macOS
- Linux
- Web
- Self-hosted
What Photon is
Photon searches for uses of images, including altered versions, to help artists identify possible copyright or license violations. Powered by DejaVuAI, it supports 1:N, N:1, and M:N similarity searches regardless of image adjustments. Listed formats include JPG, PNG, GIF, TIFF, RAW, HDR imaging, spherical dewarping, video, and floating-point images. Photon Online accepts uploads for daily searches across listed sites such as Alibaba, Etsy, Redbubble, Walmart, and Target. Photon On Prem is designed for image libraries and offers personal perpetual, personal subscription, and enterprise subscription options. The product is listed for Linux, macOS, Windows, web, and self-hosted environments. Photon says it is under 90 MB and can run from ARM processors such as Raspberry Pi through cluster computers or supercomputers. Its technology uses CPU, RAM, and cache without requiring a GPU or Tensor cores. Other listed uses include logo and product recognition, location identification, wayfinding, fake-news identification, smart manufacturing, and photo identification. Pricing is available on request.
Who it suits
Photon suits artists and organizations searching image libraries or checking where images appear online. Its deployment options also address embedded, board-level, and larger computing environments.
Good to know, and what to watch
- Finds altered image versions
- Supports multiple similarity search modes
- Handles many image and video formats
- Runs without a GPU or Tensor cores
- Offers online and on-premise options
- Pricing is available only on request
- Online searches cover a listed site set
Plans
Every plan and pricePhoton Online
Photon On Prem — Enterprise Subscription
Photon On Prem — Personal Perpetual
Photon On Prem — Personal Subscription
Pricing on request
The verdict
Photon combines broad format support with similarity searches for online and private image collections. Its CPU-based deployment range is notable, while buyers should request pricing and clarify security requirements.
What the maker publishes
- Deployment
- The technology page describes integration into computing environments, including near-sensor processing; the product page also describes board-level CPU and cluster or VM deployments. source
- GPU requirement
- The technology page says DejaVuAI relies on CPU, RAM, and cache and does not need a GPU or Tensor cores. source
- Image formats
- The site lists JPG, PNG, GIF, TIFF, RAW, HDR imaging, spherical dewarping, video, and floating point image types as supported. source
- Local and enterprise use
- Photon On Prem is described for searching image libraries, with personal perpetual, personal subscription, and enterprise subscription options. source
- Matching details
- The on-prem page says Photon automatically identifies scale and rotation of matched images and provides image registration. source
- Online searches
- Photon Online lets users upload images for daily searches against a listed set of known sites, including Alibaba, Etsy, Redbubble, Walmart, and Target. source
- Purpose
- Photon searches for uses of images, including altered versions, to help artists identify possible copyright or license violations. source
- Resource requirements
- The product page says Photon is under 90 MB and can run on CPUs ranging from an ARM processor such as Raspberry Pi to cluster computers or supercomputers. source
- Search modes
- Powered by DejaVuAI®, Photon supports 1:N, N:1, and M:N reverse image searches regardless of image adjustments. source
- Security and compliance
- The Photon pages opened do not state specific security certifications or compliance attestations. source
- Support
- Photon provides a contact form asking for the user's name, company, email address, and project details to submit an inquiry. source
- Use cases
- The site lists artist copyright protection, logo and product recognition, location identification, wayfinding, fake news identification, smart manufacturing, and photo identification as possible uses. source
Checked against the maker's pages on 4 Oct 2026.