Cadryl
Cadryl, formerly Vision Dataset Studio: an Android studio for preparing images, reviewing proposed annotations, exporting datasets and training a copy of a compatible LiteRT model.
Open the public sourceCadryl, formerly Vision Dataset Studio: an Android studio for preparing images, reviewing proposed annotations, exporting datasets and training a copy of a compatible LiteRT model. The public scope separates local dataset production, human validation, exports and the limits of on-device learning.
- Repositories
- 1
- Language
- Kotlin
- Activity
- Oct 01, 26
- Stars
- 1
- Source
- GitHub
- Status
- Connected source
Documentation and scope
Structured content from the public source documentation.
Overview
Cadryl, formerly Vision Dataset Studio: an Android studio for preparing images, reviewing proposed annotations, exporting datasets and training a copy of a compatible LiteRT model.
The public scope separates local dataset production, human validation, exports and the limits of on-device learning.
Documented scope
- Local or Hugging Face import; manual work available without a model
- Guided three-step configuration and verified model compatibility
- Pre-annotation followed by human correction and validation
- JSONL, COCO, YOLO, WebDataset and vision-language exports
- Optional learning disabled by default; no weights in the APK
- 4.2.0-rc7 prerelease; physical ARM 16 KB phone acceptance remains open
Place in the catalogue
Cadryl has a dedicated record connected to its current source.
Provenance and limits
The information is linked to the stated public source. Metrics reflect a reference snapshot and may change between synchronizations.
A record traceable to its source
Editorial information, metadata and documentation remain connected to the original public repository.
- Public identifier
- Unicorn Who Dev/Cadryl
- Latest activity
- October 1, 2026
Every source consolidated in this record
Public sources remain individually accessible. Private components are named without exposing their repository, revision or content.