wangjian169 Figure Analysis (Vision-Language Model)
Use a Vision-Language Model (VLM) to automatically analyze generated figures and produce structured text interpretations. The backend uses Qwen3-VL via the DashScope API.
Usage
python skills/figure_analysis_skill/Scripts/figure_analysis.py \
--task "Sensitivity results: imperv (mu*=1.03), zero_imperv (mu*=0.87), slope (mu*=0.36). Calibration NSE=0.948, Validation NSE=0.820. Analyze the figures: 1) fit quality in calibration and validation events; 2) subcatchment-level insights from the parameter bar chart (which subcatchments changed most, spatial heterogeneity); 3) overall calibration assessment." \
--figures <fig1.png> <fig2.png>
| Parameter | Description | Default |
|---|---|---|
| --task | Natural language task description for the VLM | (required) |
| --figures | Paths to image files to analyze (one or more) | (required) |
Output
Prints the VLM's text analysis response to stdout.