LLMog
An open-source automated data-annotation framework based on large language models, specializing in object grounding. It supports automatic annotation of image datasets and reclassification of YOLO-format datasets, and can run locally to protect data privacy.
Timeline (last 90 days)
LLMog is an open-source framework focused on object grounding that uses large models to recognize and locate targets in images and automatically generate annotation results.
LLMog was open-sourced by Reddit user mohamed-em2m.
LLMog is primarily aimed at two scenarios: automatic dataset annotation (Auto Annotation) and YOLO dataset reclassification (Reclassification).
LLMog also supports connecting to external APIs to use cloud services.
The author provides two online environments, Google Colab and Kaggle Notebook, for users to try LLMog.
Bounding boxes generated by large models are often semantically correct but exhibit deviations in pixel-level precision, and using them directly for training may introduce noisy labels.
For specialized domains such as medical imaging and industrial inspection, general-purpose large models may not perform as well as purpose-trained detection models.
All Facts (7)
LLMog is an open-source framework focused on object grounding that uses large models to recognize and locate targets in images and automatically generate annotation results.
50%UnverifiedLLMog was open-sourced by Reddit user mohamed-em2m.
50%UnverifiedLLMog is primarily aimed at two scenarios: automatic dataset annotation (Auto Annotation) and YOLO dataset reclassification (Reclassification).
50%UnverifiedLLMog also supports connecting to external APIs to use cloud services.
50%UnverifiedThe author provides two online environments, Google Colab and Kaggle Notebook, for users to try LLMog.
50%UnverifiedBounding boxes generated by large models are often semantically correct but exhibit deviations in pixel-level precision, and using them directly for training may introduce noisy labels.
50%UnverifiedFor specialized domains such as medical imaging and industrial inspection, general-purpose large models may not perform as well as purpose-trained detection models.
50%