Feature availability 🛈
⭕ Personal ✔️ Suite
AI-powered annotation in Connecter is a powerful time-saving tool, but it’s important to understand its limitations to set realistic expectations.
How AI analyzes previews
The annotation metadata is generated based solely on the visual previews of assets – not their actual content – therefore, you should look out for the following:
- Mismatched previews: If a preview doesn’t accurately represent the asset, the annotation metadata may also be irrelevant. For example, placeholder previews with generic thumbnails may result in irrelevant or misleading annotation metadata.
- Poor-quality previews: Small or low-resolution previews, especially for complex scenes, can result in incomplete or inaccurate annotation metadata.
Challenges with specific assets
While common objects and scenes yield reliable results, niche or highly specific assets may be misidentified due to their uniqueness or visual ambiguity. That is a natural limitation of AI, which performs best with recognizable patterns and objects.
Optimizing results
- Use high-quality previews: Ensure previews are clear and representative of the asset.
- Leverage custom previews: Replace inadequate embedded previews with custom ones for better accuracy.
- Review annotation metadata: For unique assets, review the generated metadata and adjust tags or labels as needed.
By understanding these factors, you can achieve more consistent and relevant results while minimizing potential frustrations.