pytorch / pytorch/vision

Wrong prediction when position of object is changed

Open
#4,198 4 comments 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

Dominant language
Python
Stars
17.9k
Forks
7.3k
Avg merge
1d 15h
Merged PRs (30d)
13

Description

I am doing an object detection on set of images. I am finetuning using the pytorch Object detection tutorial with my own set of images. I have 100-200 images. And 22 classes.

What I notice is that the model predicts well as long as the specific object is in a set position. For instance if object 1 is in position1, it predicts it as object 1. But when object 1 is moved to position2, it predicts it as object2. Object 1 and Object 2 look very similar except for few minute changes.

Training data: It is difficult for me to create a training data for 22 classes, with each class in 22 positions (Plus consider combinations of negative data, etc.). Is there a way to solve this?

Contributor guide

Open the contributing guide

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

No repository file, test, or entry point is identified. The report describes position-dependent predictions in a custom 22-class object-detection training setup; further reproduction details would be needed before defining a repository change or a verifiable completion condition.

Written by the indexing model from the issue text.

Assessment

Tech stack
python, pytorch
Domain
computer-vision, machine-learning
Issue type
Bug
Difficulty
5/5
Estimated time
Over a week
Activity status
Stale
Clarity
Needs clarification
Newbie friendliness
15/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.