facebookresearch / facebookresearch/detectron2
random init cfg file for VitDet
Open
enhancement
- Dominant language
- Python
- Stars
- 34.7k
- Forks
- 7.9k
- PR merge metrics
- No merged PRs in 30d
Description
In vitdet paper, there is some result for random init vitdet(without MAE pretraining). however, when i try to train a random init vitdet, the result of it is much lower than the paper result. 46.1 vs 48.1. So what is the cfg for random init vitdet?
Contributor guide
Research direction
The issue names no repository file, test, or entry point. Start by locating the VitDet training configuration and comparing its random-initialization settings with the paper’s experiment; done means identifying a configuration that reproduces the reported 48.1 result instead of 46.1.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- python
- Domain
- computer-vision, machine-learning
- Issue type
- Bug
- Difficulty
- 4/5
- Estimated time
- 3-5 days
- Activity status
- Stale
- Clarity
- Needs clarification
- Newbie friendliness
- 25/100