TruFor tool does not actually integrate the TruFor model

[Gooichand]

Issue Summary

TruForWidget presents itself as TruFor integration but uses a simple edge-density heuristic

Description

gui/trufor.py docstring and comments state that this file "implements a tool for integrating TruFor into Sherloq" and references the TruFor CVPR 2023 paper and repository.
However, the implementation in TruForWorker.run only checks for PyTorch availability, converts the image to grayscale, computes Canny edges, applies a simple smoothing kernel, and derives a "manipulation probability" from mean edge density.
It does not load or run the actual TruFor model or any learned weights from the referenced project.

Steps to Reproduce

1. Read the documentation block at the top of gui/trufor.py.
2. Inspect the TruForWorker.run method to see the actual analysis pipeline.
3. Compare with the original TruFor model architecture and usage.

Expected Result

The TruFor tool should either integrate the real TruFor model or clearly state that it is a heuristic "TruFor-inspired" analysis, not the official model.

Actual Result

The UI and docstring strongly imply TruFor integration, but the code implements a simple edge-based heuristic unrelated to the published TruFor network.

Affected Files

• gui/trufor.py

Suggested Fix

Either (a) add real TruFor model support (downloading weights, running inference, etc.), or (b) rename and reword the tool and documentation to avoid implying official TruFor integration (e.g., "AI-inspired heuristic analysis").

[diksha78dev]

Hi, I am OSCG'26 contributor and i would like to work on this issue, can you assign this issue to me?

[Gooichand]

@Diksha78-bot after completing issue which are assigned to you we will see about this issue

[magic-peach]

@Gooichand can i work on this

[Gooichand]

@magic-peach already issue assign to you after that i wll update new issue

[Aditya8369]

@Gooichand may I work on it

[Gooichand]

@magic-peach @Aditya8369 @diksha78dev pls comments your plan of action ao i can know are you willing to wokr on this issue or not

[Tech-Psycho95]

Hi @Gooichand , I reviewed gui/trufor.py and noticed that TruForWorker.run currently implements an edge-density heuristic rather than the published TruFor model.

Could you clarify whether this heuristic is intended as a lightweight substitute, or if full TruFor integration is planned?

I’d be happy to contribute in either direction — either by helping integrate the actual TruFor model (weights + inference pipeline) or by updating the documentation/UI to reflect that this is a heuristic “TruFor-inspired” analysis. Let me know which path aligns best with the project goals.

[Gooichand]

@Tech-Psycho95 Thank you for the thorough analysis and thoughtful question! 👏

You've identified the issue perfectly. The current implementation is indeed a heuristic edge-density approach, not the actual TruFor model from the CVPR 2023 paper.

🎯 My Decision: Option B - Update Documentation (Recommended for OSCG)

Here's my reasoning:

Why Documentation Update:

1. Realistic Scope: Full TruFor integration requires significant ML infrastructure (model weights, GPU handling, inference pipeline, dependencies management) - this is quite complex for an OSCG timeline
2. Honest Communication: Users deserve accurate information about what the tool does
3. Immediate Value: Fixing misleading documentation improves user trust right away
4. Future-Proof: Leaves door open for ML enhancement later without blocking current progress

📋 What I Need:

Documentation/Refactoring Task:

• Rename the tool to accurately reflect its functionality (e.g., "Edge-Based Manipulation Detection" or "Heuristic Image Analysis")
• Update all docstrings in gui/trufor.py to remove TruFor paper references
• Modify UI labels/descriptions to clarify this is a lightweight heuristic
• Add a comment block explaining the edge-density algorithm used
• Update any user-facing text that implies TruFor model usage

Optional Enhancement:

• If you have bandwidth, add a TODO/comment noting that future enhancement could include actual TruFor model integration

✅ Next Steps:

If you're comfortable with this approach:

1. Fork the repository
2. Create a branch (e.g., fix/trufor-documentation)
3. Make the documentation/naming updates
4. Submit a PR with clear description of changes

This is a valuable contribution that improves project integrity and user experience!

However, if you're specifically interested in the full ML integration challenge and have experience with PyTorch model integration, we can discuss that path too - but it would be a larger, multi-PR effort.

Let me know which direction you prefer! 🚀

[Tech-Psycho95]

@Gooichand Thank you for the clear guidance and for outlining the reasoning behind Option B. I completely agree that updating the documentation and naming is the most practical step right now, and it will immediately improve clarity for users.

I’ll proceed with the documentation/refactoring task as suggested:

• Rename the tool to reflect its heuristic nature

• Update docstrings and remove references to the TruFor paper

• Adjust UI labels/descriptions for transparency

• Add a comment block explaining the edge-density algorithm

• Include a TODO note for potential future TruFor integration

I’ll create a branch (fix/trufor-documentation) and submit a PR with a clear description of the changes. Looking forward to your feedback once it’s up! 🚀

[Gooichand]

@Tech-Psycho95 go ahead