Analyze the image with MLLM or online APIs
to run the analyzer: (with agiqa-3k
dataset in ./data
folder)
python analyzer.py --dataset agiqa-3k --provider minicpmv2.5 --gpu 2 --prompt ./prompt_mosq_n_mosa.template --data_path /home/user/data -o minicpm
then a file in ./data/agiqa-3k
named minicpm_analysis_output.csv
will be created. which can be process later by other purpose, such as fine-tuning a MLLM.