[Feature Request] Optionally save ImageFolder information
Nobody has claimed this yet.
- Dominant language
- Python
- Stars
- 17.9k
- Forks
- 7.3k
- Avg merge
- 1d 15h
- Merged PRs (30d)
- 13
Description
Currently ImageFolder indexes a directory when the object is created. For some datasets (e.g., Places365, ImageNet, etc.) this can take several minutes. In the case of ImageNet-22K, this can take over 30 minutes on an SSD (albeit on an NFS system). Torch saved this information, but PyTorch requires re-indexing every run. I request the option to save and restore the indexing information.
Contributor guide
First steps
- Read the whole issue, then the project's contributing guide.
- Comment on the issue to say you are picking it up — it saves two people doing the same work.
- Fork the repository and make your change on a branch.
- Open a pull request that references the issue number.
Research direction
Start by locating the ImageFolder implementation and tracing where directory indexing occurs during object creation. Review how the dataset currently represents its indexed information and any existing serialization patterns. Done means an optional save-and-restore path avoids re-indexing on later runs while preserving current behavior when it is not enabled.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- python
- Domain
- computer-vision
- Issue type
- Feature
- Difficulty
- 4/5
- Estimated time
- 3-5 days
- Activity status
- Stale
- Clarity
- Mostly clear
- Newbie friendliness
- 35/100