Re-analyzing images: improving the automatic analysis
Learn when and how to re-analyze images in Polimake to get improved tags and descriptions when the initial analysis wasn't accurate.
The team behind Polimake. We explore the intersection of technology, creativity, and automation.
Although automatic AI analysis in Polimake is very accurate, sometimes the initial analysis may not fully capture the content of an image, or it may benefit from a more detailed analysis. The re-analysis feature lets you request a new analysis to get improved tags and descriptions.
This guide will explain when to re-analyze images, how to do it, and what to expect from the process.
What is re-analyzing?
Re-analyzing is requesting a new AI analysis for an image that was previously analyzed. This:
- Generates new tags based on the most recent analysis
- Improves descriptions if the initial analysis was incomplete
- Updates metadata with more accurate information
- Can detect elements that were initially overlooked
It's useful when the initial analysis wasn't accurate or detailed enough.
When to re-analyze
Incomplete analysis
Re-analyze when:
- Important tags are missing that should be present
- The description is too generic or unhelpful
- Clearly visible elements weren't detected
- The tags don't reflect the actual content of the image "
" Example: You have a photo of a specific product, but the analysis only detected "object" without identifying the product type.
Improvements in the AI model
If Polimake updates its AI model:
- New analyses can be more accurate
- You can get better information with improved technology
- More detailed tags and more complete descriptions
Example: The AI model is updated and can now better identify certain types of products or scenes.
Industry-specific content
For images with very specific content:
- The initial analysis may use general terminology
- Re-analyzing can help, although you may need to add tags manually
- Useful for trying to get better detection
Note: For content very specific to your industry, you'll likely need to add tags manually anyway.
Important images
For images you use frequently:
- It's worth it to ensure they have the best possible analysis
- Accurate tags improve future search
- Complete descriptions help you understand the content quickly
How to re-analyze
From the detail view
To re-analyze an image:
- Open the image in detail view (click the image)
- Go to the "Analysis" tab or look for the "Re-analyze" button
- Click "Re-analyze" or "Re-analyze with AI"
- Confirm the action if prompted
- Wait for the process to complete
Button location: The "Re-analyze" button is typically in the Analysis tab or in the action bar of the detail view.
Re-analysis process
During re-analysis:
- The image is sent to the AI service
- It is processed with the most recent model
- New tags and a description are generated
- The information is updated automatically
- The results appear in the detail view
Time: It typically takes a few seconds to a minute, similar to the initial analysis.
Progress indicators
While re-analyzing, you'll see:
- A loading or progress indicator
- A message indicating the analysis is in progress
- An automatic update when it completes
Note: You can keep working while it's processing. The results will update automatically when they're ready.
What to expect from re-analysis
Possible improvements
Re-analysis can improve:
- More tags detected: Additional elements identified
- More specific tags: More precise terminology
- Improved description: More complete and useful text
- Better categorization: Tags better organized by type
Variable results
Results may vary:
- Sometimes they improve significantly: New elements detected
- Sometimes they're similar: The initial analysis was already good
- Occasionally different: A slightly different focus but equally valid
Realistic expectation: There won't always be a dramatic improvement, but it's worth trying if the initial analysis wasn't satisfactory.
Existing tags
When you re-analyze:
- Previous tags are replaced with the new ones
- Manual tags you added may be kept or may need to be re-added
- Verify after re-analysis to make sure you have all the information you need
Tip: If you added important manual tags, note them before re-analyzing so you can re-add them afterward if necessary.
Best practices
Review before re-analyzing
Before re-analyzing:
- Review the current analysis to understand what's missing
- Identify what you expect from the new analysis
- Consider whether manual tags would be more efficient
Combine with manual editing
For best results:
- Re-analyze first to get an improved analysis
- Then add manual tags for information specific to your industry
- Combine both for maximum usefulness
Don't overuse re-analysis
For efficiency:
- Re-analyze only when necessary: Not all images need it
- The initial analysis is usually sufficient: For most cases
- Use it selectively: For important images or those with a poor analysis
Limitations
No guaranteed improvement
Re-analysis:
- Doesn't guarantee that the results will be better
- May give results similar to the initial analysis
- Depends on the image and the AI model
Very specific content
For very specific content:
- It may still use general terminology
- Manual tags may be necessary anyway
- It doesn't replace specific domain knowledge
Time and resources
Re-analysis:
- Takes time (seconds to minutes)
- Uses system resources
- May have limits depending on your plan
FAQ about re-analysis
How many times can I re-analyze an image?
There's typically no strict limit, but re-analyzing multiple times generally doesn't improve results significantly.
Are manual tags kept after re-analyzing?
It depends on the implementation. Some versions keep manual tags, others replace them. Verify after re-analyzing.
Is re-analyzing free?
It depends on your plan. Some plans may have limits on the number of re-analyses. Check the details of your plan.
Can I re-analyze multiple images at once?
Currently, re-analysis is for individual images. For multiple images, you need to re-analyze them one by one.
Does re-analysis use a different AI model?
It may use the most recent model available, which can be more advanced than the one used in the initial analysis if Polimake has updated its models.
What if the re-analysis is worse than the original?
You can re-analyze again, although results are generally similar or better. If you prefer, you can keep manual tags based on the original analysis.
Conclusion
Re-analyzing images is a useful tool for improving the analysis when the initial one wasn't satisfactory. Use it selectively for important images or when the initial analysis clearly didn't capture the content adequately.
Remember that for content very specific to your industry, manual tags are still valuable and often necessary. Combine re-analysis with manual tag editing for better results.
Next steps
- Identify images with an incomplete or inaccurate analysis
- Re-analyze an image to see the difference
- Compare results before and after re-analysis
- Add manual tags to supplement the analysis
- Establish criteria for when to re-analyze in your workflow
Improve the analysis quality of your important images!