This is repository of the 6th place solution of kaggle RSNA STR Pulmonary Embolism Detection.
The discription of this solution is available here.
The prediction notebook in the competition is available here.

Google Cloud Platform

• SMP Debian 4.9.210-1 (2020-01-20)
• n1-standard-16 (vCPU x 16, memory 60 GB)
• 1 x NVIDIA Tesla P100

• Python 3.7.6
• CUDA 10.1
• cuddn 7.6.3
• gdcm 2.8.9
• numpy==1.19.1
• pandas==1.1.1
• matplotlib==3.2.1
• opencv-python==4.3.0.36
• pydicom==2.0.0
• scikit-learn==0.23.1
• torch==1.6.0+cu101
• torchvision==0.7.0+cu101
• timm==0.1.26
• albumentations==0.4.5

Download the competition dataset and place them in input/orig/.
In case you use pretrained weights, download the weights and place them to models/.

Run all the cells of notebook/preprocess.ipynb.
Run all the cells of notebook/train_stage1.ipynb.
run all cells of notebook/train_stage2.ipynb.
Rewrite the line 9 of the 5th cell of notebook/train_stage1.ipynb to MODEL_NAME = 'b2' and run all the cells.
Rewrite the line 8 of the 5th cell of notebook/train_stage1.ipynb to MODEL_NAME = 'b2' and run all the cells.
Run all the cells of notebook/postprocess.ipynb.
Run all the cells of notebook/predict.ipynb.

Rewrite the line 12-15 of the 3rd cell of notebook/predict.ipynb to

weight_dir_b0_1 = "../models/b0_stage1"
weight_dir_b0_2 = "../models/b0_stage2"
weight_dir_b2_1 = "../models/b2_stage1"
weight_dir_b2_2 = "../models/b2_stage2"

and run all the cells.