Detect-Anything(designed for practical application) is a intelligent object detection annotation helper who combines the capabilities of multiple foundation models including Segment Anything, BLIPv2, Flan-T5, GPT2, and even ChatBot and GPT4.
We want capture more precise texonomy for detection and instance segmentation by exploiting these AI-Large models.
Sparing no effort to for exploitation!!!
Coming soon...
- Saving COCO format
- several threshold interface
# Example: without chatgpt
export CUDA_VISIBLE_DEVICES=0
python pipeline_DA.py \
--data_dir
--out_dir