endoscopic video recognize vessels fine-tune with sam2

I want to identify blood vessels in laparoscopic videos and plan to use SAM2 to train on my own laparoscopic dataset. I need to automatically recognize and segment more than ten types of vessels, such as the left gastric artery and right gastric vein, and I’d like to avoid using prompts if possible. Their image features appear quite similar, and I’m concerned whether fine-tuning on SAM2 can meet our expectations. Do you have any good suggestions?