Skip to content

Latest commit

 

History

History
35 lines (24 loc) · 1.84 KB

readme.md

File metadata and controls

35 lines (24 loc) · 1.84 KB

Stage 3 Dataset Generation

Generate fossil caption data for stage 3 training.

python run.py --module feat_recognize.generate_dataset \
    --save_data_path dataset/ \
    --desc_prompt_dir feat_recognize/prompt_icl.txt \
    --desc_llm qwen25-14 \
    --desc_batchsize 1 \
    --paraphrase_prompt_dir data/caption/paraphrase_stage3.txt \
    --caption_llm api-deepseek-chat \
    --caption_batchsize 1

The generation process contains 3 parts:

  1. Recognize visual features of the fossil images. The results are in {save_data_path}/instructions.jsonl.
  2. Replace numerical information in original fossil description. The results are in {save_data_path}/stage3.jsonl
  3. Paraphrase the description to enhance diversity of description. The results are in {save_data_path}/stage3_paraphrase.jsonl

--desc_prompt_dir is the prompt text file for numerical information replacement. --desc_llm specify the llm for replacement. Similarly, --caption_prompt_dir and --caption_llm specify the prompt text file and LLM for generating diverse paraphrased descriptions.

Feature recognition for specific part of a fossil image.

Initial Chamber

The detect_initial_chamber function in initial_chamber.py uses cv2.houghcircle to detect the circle in the central region of the image.

Volutions

The VolutionCounter class in volution_counter.py is designed to detect and measure the volutions with a "adsorb-scan" strategy. Call the count_volutions method with the input image to get the volutions-related features, it also returns whether the initial chamber was detected with a high confidence level.

Usage Example

The recognize_feature function in recognize.py is an example of feature recognition. It first recognizes the volutions in the fossil image, and try to detect the initial chamber twice with different confidence level.