search

Luffy03 / Large-Scale-Medical

[CVPR 2024 Extension] 160K volumes (42M slices) datasets, new segmentation datasets, 31M-1.2B pre-trained models, various pre-training recipes, 50+ downstream tasks implementation
Apache License 2.0
33 stars 5 forks source link

Image Segmention #3

Open NirvanaChan opened 3 hours ago

NirvanaChan commented 3 hours ago

Can I use this pre-trained model to segment unlabeled images? Does it support point prompts or text prompts？

Luffy03 commented 2 hours ago

Hi, thanks for your attention to our work and sorry for my late reply. Sure you can use our models to segment unlabeled images, you can use the template as test.py to predict the images, and the output format as VoComni.json. But if you want to predict some specific types of tumors, I recommend you to finetune with some downstream datasets since the pre-trainind models also need finetuning. For point prompts or text prompts, I apologize for any confusion. This work is intended for backbone pretraining and is not compatible with SAM-based models, thus ours do not support prompts.