SIIM-COVID19-Detection

Source code of the 1st place solution for SIIM-FISABIO-RSNA COVID-19 Detection Challenge.

1.INSTALLATION

Ubuntu 18.04.5 LTS
CUDA 10.2
Python 3.7.9
python packages are detailed separately in requirements.txt

$ conda create -n envs python=3.7.9
$ conda activate envs
$ conda install -c conda-forge gdcm
$ pip install -r requirements.txt
$ pip install git+https://github.com/ildoonet/pytorch-gradual-warmup-lr.git

2.DATASET

2.1 SIIM COVID 19 DATASET

download competition dataset at link then extract to ./dataset/siim-covid19-detection

$ cd src/prepare
$ python dicom2image_siim.py
$ python kfold_split.py
$ prepare_siim_annotation.py                        # effdet and yolo format
$ cp -r ../../dataset/siim-covid19-detection/images ../../dataset/lung_crop/.
$ python prepare_siim_lung_crop_annotation.py

2.2 EXTERNAL DATASET

download pneumothorax dataset at link then extract to ./dataset/external_dataset/pneumothorax/dicoms
download pneumonia dataset at link then extract to ./dataset/external_dataset/rsna-pneumonia-detection-challenge/dicoms
download vinbigdata dataset at link then extract to ./dataset/external_dataset/vinbigdata/dicoms
download chest14 dataset at link then extract to ./dataset/external_dataset/chest14/images
download chexpert high-resolution dataset at link or then extract to ./dataset/external_dataset/chexpert/train
download padchest dataset at link or then extract to ./dataset/external_dataset/padchest/images
_{most of the images in bimcv and ricord overlap with siim covid trainset and testset. To avoid data-leak when training, I didn't use them. You can use script}

$ cd src/prepare
$ python dicom2image_pneumothorax.py
$ python dicom2image_pneumonia.py
$ python prepare_pneumonia_annotation.py      # effdet and yolo format
$ python dicom2image_vinbigdata.py
$ python prepare_vinbigdata.py
$ python refine_data.py                       # remove unused file in chexpert + chest14 + padchest dataset
$ python resize_padchest_pneumothorax.py

dataset structure should be ./dataset/dataset_structure.txt

3.SOLUTION SUMMARY

4.TRAIN MODEL

4.1 Classification

4.1.1 Multi head classification + segmentation

Stage1

$ cd src/classification_aux
$ bash train_chexpert_chest14.sh              #Pretrain backbone on chexpert + chest14
$ bash train_rsnapneu.sh                      #Pretrain rsna_pneumonia
$ bash train_siim.sh                          #Train siim covid19

Stage2: Generate soft-label for classification head and mask for segmentation head.
Output: soft-label in ./pseudo_csv/[source].csv and public test masks in ./prediction_mask/public_test/masks

$ bash generate_pseudo_label.sh [checkpoints_dir]

Stage3: Train model on trainset + public testset, load checkpoint from previous round

$ bash train_pseudo.sh [previous_checkpoints_dir] [new_checkpoints_dir]

Rounds of pseudo labeling (stage2) and retraining (stage3) were repeated until the score on public LB didn't improve.

For final submission

$ bash generate_pseudo_label.sh checkpoints_v3
$ bash train_pseudo.sh checkpoints_v3 checkpoints_v4

For evaluation

$ CUDA_VISIBLE_DEVICES=0 python evaluate.py --cfg configs/xxx.yaml --num_tta xxx

[email protected] 4 classes: negative, typical, indeterminate, atypical

	SeR152-Unet	EB5-Deeplab	EB6-Linknet	EB7-Unet++	Ensemble
w/o TTA/8TTA	0.575/0.584	0.583/0.592	0.580/0.587	0.589/0.595	0.595/0.598

_{8TTA: (orig, center-crop 80%)x(None, hflip, vflip, hflip & vflip). In final submission, I use 4.1.2 lung detector instead of center-crop 80%}

4.1.2 Lung Detector-YoloV5

I annotated the train data(6334 images) using LabelImg and built a lung localizer. I noticed that increasing input image size improves the modeling performance and lung detector helps the model to reduce background noise.

$ cd src/detection_lung_yolov5
$ cd weights && bash download_coco_weights.sh && cd ..
$ bash train.sh

	Fold0	Fold1	Fold2	Fold3	Fold4	Average
[email protected]:0.95	0.921	0.931	0.926	0.923	0.922	0.9246
[email protected]	0.997	0.998	0.997	0.996	0.998	0.9972

4.2 Opacity Detection

Rounds of pseudo labeling (stage2) and retraining (stage3) were repeated until the score on public LB didn't improve.

4.2.1 YoloV5x6 768

Stage1:

$ cd src/detection_yolov5
$ cd weights && bash download_coco_weights.sh && cd ..
$ bash train_rsnapneu.sh          #pretrain with rsna_pneumonia
$ bash train_siim.sh              #train with siim covid19 dataset, load rsna_pneumonia checkpoint

Stage2: Generate pseudo label (boxes)

$ bash generate_pseudo_label.sh

Jump to step 4.2.4 Ensembling + Pseudo labeling

Stage3:

$ bash warmup_ext_dataset.sh      #train with pseudo labeling (public-test, padchest, pneumothorax, vin) + rsna_pneumonia
$ bash train_final.sh             #train siim covid19 boxes, load warmup checkpoint

4.2.2 EfficientDet D7 768

Stage1:

$ cd src/detection_efffdet
$ bash train_rsnapneu.sh          #pretrain with rsna_pneumonia
$ bash train_siim.sh              #train with siim covid19 dataset, load rsna_pneumonia checkpoint

Stage2: Generate pseudo label (boxes)

$ bash generate_pseudo_label.sh

Jump to step 4.2.4 Ensembling + Pseudo labeling

Stage3:

$ bash warmup_ext_dataset.sh      #train with pseudo labeling (public-test, padchest, pneumothorax, vin) + rsna_pneumonia
$ bash train_final.sh             #train siim covid19, load warmup checkpoint

4.2.3 FasterRCNN FPN 768 & 1024

Stage1: train backbone of model with chexpert + chest14 -> train model with rsna pneummonia -> train model with siim, load rsna pneumonia checkpoint

$ cd src/detection_fasterrcnn
$ CUDA_VISIBLE_DEVICES=0,1,2,3 python train_chexpert_chest14.py --steps 0 1 --cfg configs/resnet200d.yaml
$ CUDA_VISIBLE_DEVICES=0,1,2,3 python train_chexpert_chest14.py --steps 0 1 --cfg configs/resnet101d.yaml
$ CUDA_VISIBLE_DEVICES=0 python train_rsnapneu.py --cfg configs/resnet200d.yaml
$ CUDA_VISIBLE_DEVICES=0 python train_rsnapneu.py --cfg configs/resnet101d.yaml
$ CUDA_VISIBLE_DEVICES=0 python train_siim.py --cfg configs/resnet200d.yaml --folds 0 1 2 3 4 --SEED 123
$ CUDA_VISIBLE_DEVICES=0 python train_siim.py --cfg configs/resnet101d.yaml --folds 0 1 2 3 4 --SEED 123

_{Note: Change SEED if training script runs into issue related to augmentation (boundingbox area=0) and comment/uncomment the following code if training script runs into issue related to resource limit}

import resource
rlimit = resource.getrlimit(resource.RLIMIT_NOFILE)
resource.setrlimit(resource.RLIMIT_NOFILE, (8192, rlimit[1]))

Stage2: Generate pseudo label (boxes)

$ bash generate_pseudo_label.sh

Jump to step 4.2.4 Ensembling + Pseudo labeling

Stage3:

$ CUDA_VISIBLE_DEVICES=0 python warmup_ext_dataset.py --cfg configs/resnet200d.yaml
$ CUDA_VISIBLE_DEVICES=0 python warmup_ext_dataset.py --cfg configs/resnet101d.yaml
$ CUDA_VISIBLE_DEVICES=0 python train_final.py --cfg configs/resnet200d.yaml
$ CUDA_VISIBLE_DEVICES=0 python train_final.py --cfg configs/resnet101d.yaml

4.2.4 Ensembling + Pseudo labeling

Keep images that meet the conditions: negative prediction < 0.3 and maximum of (typical, indeterminate, atypical) predicion > 0.7. Then choose 2 boxes with the highest confidence as pseudo labels for each image.

_{Note: This step requires at least 128 GB of RAM}

$ cd ./src/detection_make_pseudo
$ python make_pseudo.py
$ python make_annotation.py

4.2.5 Detection Performance

	YoloV5x6 768	EffdetD7 768	F-RCNN R200 768	F-RCNN R101 1024
[email protected] TTA	0.580	0.594	0.592	0.596

Final result: Public LB/Private LB: 0.658/0.635

5.FINAL KERNEL

siim-covid19-2021
demo notebook to visualize output of models

6.AWESOME RESOURCES

Pytorch ✨
PyTorch Image Models ✨
Segmentation models ✨
EfficientDet ✨
YoloV5 ✨
FasterRCNN FPN ✨
Albumentations ✨
Weighted boxes fusion ✨

1st place solution for SIIM-FISABIO-RSNA COVID-19 Detection Challenge

Related tags

Overview

SIIM-COVID19-Detection

1.INSTALLATION

2.DATASET

2.1 SIIM COVID 19 DATASET

2.2 EXTERNAL DATASET

3.SOLUTION SUMMARY

4.TRAIN MODEL

4.1 Classification

4.1.1 Multi head classification + segmentation

4.1.2 Lung Detector-YoloV5

4.2 Opacity Detection

4.2.1 YoloV5x6 768

4.2.2 EfficientDet D7 768

4.2.3 FasterRCNN FPN 768 & 1024

4.2.4 Ensembling + Pseudo labeling

4.2.5 Detection Performance

5.FINAL KERNEL

6.AWESOME RESOURCES

Owner

Nguyen Ba Dung

([email protected]) Boosting Co-teaching with Compression Regularization for Label Noise

Balabobapy - Using artificial intelligence algorithms to continue the text

Thresholding-and-masking-using-OpenCV - Image Thresholding is used for image segmentation

RRD: Rotation-Sensitive Regression for Oriented Scene Text Detection

Code release for Hu et al., Learning to Segment Every Thing. in CVPR, 2018.

Fast image augmentation library and easy to use wrapper around other libraries. Documentation: https://albumentations.ai/docs/ Paper about library: https://www.mdpi.com/2078-2489/11/2/125

A python program to block out your face

Image augmentation for machine learning experiments.

Scene text detection and recognition based on Extremal Region(ER)

Code for the AAAI 2018 publication "SEE: Towards Semi-Supervised End-to-End Scene Text Recognition"

Generate text images for training deep learning ocr model

Memory tests solver with using OpenCV

OCR, Object Detection, Number Plate, Real Time

Indonesian ID Card OCR using tesseract OCR

Image Recognition Model Generator

EQFace: An implementation of EQFace: A Simple Explicit Quality Network for Face Recognition

Reference Code for AAAI-20 paper "Multi-Stage Self-Supervised Learning for Graph Convolutional Networks on Graphs with Few Labels"

GDB python tool to pretty print and debug c++ xtensor containers

A facial recognition device is a device that takes an image or a video of a human face and compares it to another image faces in a database.