ML_project

PKU machine learning project 2023 spring

Installation

follow installation guidance in segment-anything project.

• python 3.7
• pytorch 1.7.1
• CUDA capable GPU (one is enough)

Preparation

download preprocessed CT data and fine-tuned model weights from this link

• place model weights in ./pre_train/

• place data for train and test in ./BTCV/

Task1

Zero-shot performance on test set

cd Task2

python main.py --promt_type box --test

You may also define:

• --device_id id.
• --seed random seed.
• --center_point choose center point prompt or random choose point prompt
• --ckpt_path path/to/your/sam_model_decoder_weights

you can also modify these options in Task2/config/cfg.yaml

Task2

Fine-tune on training set

Task3