onnx / onnx/models

ArcFace input clarification

Open
#551 0 comments 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

question
Dominant language
Jupyter Notebook
Stars
9.8k
Forks
1.6k
PR merge metrics
No merged PRs in 30d

Description

Ask a Question

Question

I am trying to use ArcFace, and I am not sure how to format the input image into the model's 'data' tensor;

this line of code suggests that RGB values of range 0-255 must be translated to values between -0.5 to +0.5

But then, looking at the values stored inside the input_0.pb of the test data are in the range of about -35 to +35

Further information

Trying to use this from C# ... is there any existing example using ImageSharp?

Is this issue related to a specific model?

arcfaceresnet100-8.onnx

Notes

It could be nice if all the models of the model zoo would have a small section describing the RGB to Tensor conversion, specially if it's not a plain normalization and do require some odd transformations like mean/stddev

Contributor guide

No contributing guide indexed for this repository

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

Research direction

Start with vision/body_analysis/arcface/dependencies/helper.py at the referenced line and compare its preprocessing with the input_0.pb file in the linked test data. Clarify the expected ArcFace tensor conversion and whether an ImageSharp example exists, then document the preprocessing guidance for this model or the model zoo.

Written by the indexing model from the issue text.

Assessment

Tech stack
csharp
Domain
computer-vision, documentation, machine-learning
Issue type
Documentation
Difficulty
3/5
Estimated time
1-2 days
Activity status
Stale
Clarity
Needs clarification
Newbie friendliness
35/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.