DMMPT:Dynamic Multi-Modal Prompt Tuning: Boost Few-shot Learning with VLM-Guided Point Cloud Models

Official implementation of Dynamic Multi-Modal Prompt Tuning: Boost Few-shot Learning with VLM-Guided Point Cloud Models

Full Paper Here

We have been accepted by ECAI2024!
We have upload raw code,it is ok to run after putting your desired pre-trained encoders' checkpoint in folder /checkpoint and use .sh files to run our model. We will polish code writing and write a detailed guide in the future, we will upload our model checkpoint as well.

Name		Name	Last commit message	Last commit date
Latest commit History 11 Commits
datasets		datasets
m2932		m2932
models		models
scripts		scripts
utils		utils
Boost_Few-shot_Learning_with_VLM-Guided_Point_Cloud_Models(ECAI2024_m2932).pdf		Boost_Few-shot_Learning_with_VLM-Guided_Point_Cloud_Models(ECAI2024_m2932).pdf
FAIA-392-FAIA240559.pdf		FAIA-392-FAIA240559.pdf
README.md		README.md
batchtime.py		batchtime.py
mem.png		mem.png
pre-process_scanobjectnn.py		pre-process_scanobjectnn.py
preprocess_mn40.py		preprocess_mn40.py
prompt.py		prompt.py
prompt_16shot.py		prompt_16shot.py
prompt_16shot_ablation_psize.py		prompt_16shot_ablation_psize.py
prompt_16shot_imgfusion.py		prompt_16shot_imgfusion.py
prompt_Kshot.py		prompt_Kshot.py
prompt_Kshot_scanobjnn.py		prompt_Kshot_scanobjnn.py
prompt_fs_mul.py		prompt_fs_mul.py
prompt_fs_text.py		prompt_fs_text.py
run.sh		run.sh
test copy.py		test copy.py
test.py		test.py
test.sh		test.sh
test_scanobjnn.py		test_scanobjnn.py
watchdog.py		watchdog.py
watchdog.sh		watchdog.sh