ONNX-Image-Tagger
- Active · last change 6-3-2026
- ★ 1 on GitHub
README
imagetagger
Image tagging CLI using the Smiling Wolf wd-eva02-large-tagger-v3 ONNX model.
Build
Requires Delphi 11 Alexandria or 12 Athens with 64-bit compiler:
dcc64 -B imagetagger.dprNote: 64-bit build is required due to ONNX model memory requirements.
Deployment
SQLite (Statically Linked)
SQLite is statically linked into the executable via FireDAC.Phys.SQLiteWrapper.Stat.
No external sqlite3.dll is required.
ONNX Runtime DLL
The ONNX Runtime DLL (onnxruntime.dll) should be available in:
C:\Windows\System32\onnxruntime.dll(64-bit)- Or copy to the executable directory
Usage
imagetagger tag <image> [options]
imagetagger history
imagetagger clear-cache
imagetagger --help
imagetagger --versionTag an image
imagetagger tag photo.jpg
imagetagger tag photo.jpg --threshold 0.5 --top 10
imagetagger tag photo.jpg --format json
imagetagger tag photo.jpg --format csvOptions
--threshold <0.0-1.0>- Minimum confidence threshold (default: 0.35)--top <n>- Maximum number of tags to return (default: 20)--format <text|json|csv>- Output format (default: text)--db <path>- Custom database path
View history
imagetagger historyClear cached model
imagetagger clear-cacheModel
On first run, the tool downloads the wd-eva02-large-tagger-v3 onnx model from HuggingFace.
Models are cached in the models/ subdirectory next to the executable.

