Low confidence score (~0.3) for clear face images using SCRFD model

I have integrated the SCRFD face detection model into my application. However, I am encountering an issue where the model returns a very low confidence score (~0.3) even when the input image contains a clearly visible and well-defined face.

**Details:**
The input image [has:](https://drive.google.com/file/d/1vtKQcKBt9N4Lh3Tn--kX7CRvAmsvR_ED/view?usp=sharing)
A clear, frontal face
Good lighting conditions
No significant occlusion
**Expected [behavior:]**
The model should return a higher confidence score for such images
Actual behavior:
The model outputs a confidence score around 0.3, which is unexpectedly low

**Could you please help clarify:**

What could cause such low confidence scores for clear face images?
Are there recommended preprocessing steps or thresholds for optimal results?

The issue names no repository files, tests, or entry points. Start by reproducing SCRFD inference with the linked clear image and documenting the model variant, preprocessing, and threshold configuration; done means identifying the cause of the ~0.3 score or documenting the missing integration details.

Written by the indexing model from the issue text.