A recurring request from our community is the recognition of special content - for example, astronomical motifs such as nebulae, galaxies or star clusters, or certain animal and plant species. We take these requests seriously. At the same time, we would like to explain why Excire does not (yet) reliably recognize such content - and how you can help to improve it.
Why doesn't Excire recognize galaxies or certain bird species?
Excire uses AI models for automatic image analysis. These models have been trained with large, general image datasets.
Contents like:
- Sun, moon or starry sky (generic)
- Dog, cat or horse
- Human faces
are reliably recognized because they frequently appear in training data.
Special topics such as:
- M33 galaxy
- Eagle Nebula
- Bluebirds or orchid species
hardly appear there, if at all. Accordingly, Excire lacks the necessary learning examples for correct classification.
Why is retraining not easy?
Training an AI for new content is technically complex:
- Amount of data: At least 1000 high-quality photos per category (e.g. "galaxy") are required. Often significantly more is useful.
- Data quality: The images must show the target content clearly and in a variety of ways (different perspectives, lighting conditions, image composition).
- Relevance in the image: The content to be recognized must be dominant in the image - small or inconspicuous objects are unsuitable.
- Annotations / markings (for finer classifications): If, for example, bird species are to be differentiated, not only would the bird have to be in the image, it would also have to be precisely marked.
An example: For reliable recognition of the Milky Way, it is not enough to provide 1000 x the same night sky. It needs variety: different geographical locations, seasons, camera settings, exposure times, etc.
Can I train Excire myself?
No. It is currently not possible to retrain Excire yourself. This is explained in detail in the FAQ for version 2025. The AI is only trained internally by our development team.
How can I still help?
We welcome community support in the form of well-prepared image archives. These should fulfill the following criteria:
- Organized in folders, each with a category (e.g. "M42", "Open Star Cluster", "Sparrows")
- At least 1000 images per folder
- Resolution: approx. 1280 × 1280 pixels (longest side), aspect ratio remains the same
- No disturbing content in the image, clear dominance of motif
- As diverse as possible (scenarios, angles, image styles)
If you would like to provide such an archive, please contact support. We will then check together whether the data is suitable.
Why does all this take so long?
The further development of reliable AI takes time. It not only has to learn that something is "a galaxy", but also which galaxy - even in difficult lighting conditions or with image interference. In addition, we deliberately work locally and in a data protection-friendly manner, which does not allow online queries or access to global knowledge databases.
Conclusion
We understand the desire for specialized detection very well. Technically, this is possible - but not without great effort. If you would like to support us with suitable image data sets, you are very welcome.
Related Articles
How can I change my registered e-mail address?
Customers who purchased their license from our store at Fastspring on or after 04/18/2023 can change their email address themselves in their Excire user account. To do so, access the Excire user account via this link: https://id.excire.com/user/login ...
Switching from Windows to Mac (and vice versa): Create a new database in Excire Foto
When switching from a Windows computer to a Mac computer, you will need to recreate your Excire photo database and add your photos again. Unfortunately, the existing database cannot be transferred. This is mainly due to the different path conventions ...
Problems with certain Nikon RAW formats (High Efficiency Compression)
Problem: Photos saved in the camera as RAW with Nikon High Efficiency Compression are not displayed correctly in the 1:1 preview in Excire Foto. Background: Some Nikon cameras (e.g. Z6, Z7, Z8, Z9) offer the option of setting "High Efficiency ...
Excire Foto or Excire Search?
We have two different products on offer: 1. Excire Foto, our standalone version, which offers additional features such as Excire Analytics. 2. Excire Search, our plugin for Lightroom Classic, which integrates automatic keywording and advanced search ...
How can I create a backup copy of the Excire Foto database?
To protect yourself against data loss in the event of a hardware defect, it is advisable to create a backup copy of the Excire Foto database. (Of course, you should also back up the actual photos regularly. But this is not covered in this article). ...