tensorflow / tensorflow/tflite-support

Possibility of channel-by-channel image normalization when adding metadata

Open
#968 1 comment 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

Dominant language
C++
Stars
441
Forks
146
PR merge metrics
No merged PRs in 30d

Description

My model is trained on a certain mean and std:
mean = [151.2405, 119.5950, 107.8395]
std = [63.0105, 56.4570, 55.0035]
The problem is that the attempt to add these values to the metadata fails at the verification stage in mediapipe studio: "Error: UNIMPLEMENTED: Per-channel image normalization is not available.; Initialize was not ok; StartGraph failed."
Did I understand correctly that there is no possibility of channel-by-channel normalization? Are there any plans to add it? Unfortunately, if I normalize all channels with one number, the quality of my model deteriorates significantly.

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

The reported entry point is metadata verification in MediaPipe Studio, where per-channel image normalization currently fails. Start by tracing that verification and graph-start path; done means metadata with separate mean and std values for each channel verifies successfully and the graph starts.

Written by the indexing model from the issue text.

Assessment

Tech stack
cpp, tensorflow
Domain
machine-learning
Issue type
Feature
Difficulty
5/5
Estimated time
Over a week
Activity status
Stale
Clarity
Needs clarification
Newbie friendliness
25/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.