Localization of thoracic abnormalities model based on VinBigData (top 1%)

Last update: May 24, 2022

Overview

Repository contains the code for 2nd place solution of VinBigData Chest X-ray Abnormalities Detection competition. The goal of competition was to automatically localize and classify thoracic abnormalities from chest radiographs.

Kaggle forum posts about the solution:

Details

Solution consists of 3 parts. Each part is models from each team member. Predictions of each part in the end ensembled in single 2nd place submission on LeaderBoard. You can use only inference or train models from scratch.

Warning: since some of the data is hosted on Kaggle, in order to be able to download it, save your Kaggle API token to .kaggle/kaggle.json

Only inference

cd part_zfturbo
pip install -r requirements.txt
sh ./preproc.sh
sh ./inference.sh
cd ..

cd part_ivan
sh ./setup.sh
sh ./preproc.sh
sh ./inference.sh
cd ..

cd part_sergey
sh ./inference.sh
cd ..

python3 ensemble_models.py

Train

cd part_zfturbo
pip install -r requirements.txt
sh ./preproc.sh
sh ./train.sh
sh ./inference.sh
cd ..

cd part_ivan
sh ./setup.sh
sh ./preproc_train.sh
sh ./train.sh
sh ./inference.sh
cd ..

cd part_sergey
sh ./train.sh
sh ./inference.sh
cd ..

python3 ensemble_models.py

Comments

Question to sergey part, 640 or 1024 size?

Question to Sergey Part, why he first resizes to 640 for inference, but then normalizes the resultant boxes like they were for 1024*1024 pixels image?

pred_boxes, pred_scores, pred_labels = predict_for_files(weights, folder, imagenames, 640, is_TTA)
....
if len(cur_boxes) > 0:
    cur_boxes[:, [0, 2]] = (cur_boxes[:, [0, 2]] * image_width / 1024).astype(int)
    cur_boxes[:, [1, 3]] = (cur_boxes[:, [1, 3]] * image_height / 1024).astype(int)

opened by RedMoon32 2

Regarding Learning Rate

Regarding ivan part on mmdetection Can you please upload schedule_1x.py. The below error is seen. There was another solution recommended in (https://github.com/open-mmlab/mmdetection/issues/6456) Can you confirm on the same.

Traceback (most recent call last): File "/home/prashant/anaconda3/envs/kmmdet/lib/python3.7/site-packages/mmcv/utils/registry.py", line 52, in build_from_cfg return obj_cls(**args) File "/home/prashant/anaconda3/envs/kmmdet/lib/python3.7/site-packages/mmcv/runner/hooks/lr_updater.py", line 264, in init super(CosineAnnealingLrUpdaterHook, self).init(**kwargs) TypeError: init() got an unexpected keyword argument 'step'

opened by kailasdayanandan 1
No hyper-parameter yaml found

Thanks for sharing your great work.

I was trying to run the training code for yolov5, but then it gives me an error saying "AssertionError: File Not Found: data/hyp.scratch.yaml"

I can not find any scrip that generates hyp.scratch.yaml.

Am I missing something?

Thanks, Joseph

opened by shreka116 2
No MegaMix 341_healthy file

Hello! Again question about Sergey part) After inference, postprocess.py requires "../MegaMix/341_healthy.csv" file, can you please say where I can download it and why it is needed? Блин чет забыл что авторы русские и написал на английском вопрос опять)

opened by RedMoon32 1
Any pre trained weights

Thanks for posting the solution. do you have guys have trained weights that can be used for further fine tuning of different related tasks ? Regards Jaideep

opened by jaideep11061982 2

Releases(v1.0)

v1.0(Apr 12, 2021)

Source code(tar.gz)
Source code(zip)
retinanet_resnet101_500_classes_0.4986.h5(752.34 MB)
retinanet_resnet101_sqr.zip(989.96 MB)
retinanet_resnet101_sqr_removed_rads.zip(989.96 MB)
yolov5x.pt(167.96 MB)
yolo_best.zip(1129.64 MB)

Owner

GitHub Repository

Table Extraction Tool

Tree Structure - Table Extraction Fonduer has been successfully extended to perform information extraction from richly formatted data such as tables.

88 Jun 02, 2022

Awesome multilingual OCR toolkits based on PaddlePaddle （practical ultra lightweight OCR system, provide data annotation and synthesis tools, support training and deployment among server, mobile, embedded and IoT devices）

English | 简体中文 Introduction PaddleOCR aims to create multilingual, awesome, leading, and practical OCR tools that help users train better models and a

27.5k Jan 08, 2023

Rest API Written In Python To Classify NSFW Images.

✨ NSFW Classifier API ✨ Rest API Written In Python To Classify NSFW Images. Fastest Solution If you don't want to selfhost it, there's already an inst

23 Dec 30, 2022

An Implementation of the alogrithm in paper IncepText: A New Inception-Text Module with Deformable PSROI Pooling for Multi-Oriented Scene Text Detection

InceptText-Tensorflow An Implementation of the alogrithm in paper IncepText: A New Inception-Text Module with Deformable PSROI Pooling for Multi-Orien

115 Dec 12, 2022

7th place solution

SIIM-FISABIO-RSNA-COVID-19-Detection 7th place solution Validation: We used iterative-stratification with 5 folds (https://github.com/trent-b/iterativ

11 Jul 17, 2022

TextBoxes: A Fast Text Detector with a Single Deep Neural Network https://github.com/MhLiao/TextBoxes 基于SSD改进的文本检测算法，textBoxes_note记录了之前整理的笔记。

TextBoxes: A Fast Text Detector with a Single Deep Neural Network Introduction This paper presents an end-to-end trainable fast scene text detector, n

24 Apr 28, 2022

Motion Detection Squid Game with OpenCV Python

*Motion Detection Squid Game with OpenCV Python i am newbie in python. In this project I made a simple game to follow the trend about the red light gr

17 Nov 22, 2022

IMGUR5K handwriting set. It is a handwritten in-the-wild dataset, which contains challenging real world handwritten samples from different writers.The dataset is shared as a set of image urls with annotations. This code downloads the images and verifies the hash to the image to avoid data contamination.

IMGUR5K Handwriting Dataset To run the code for downloading the urls and generate corresponding annotations : Usage: python download_imgur5k.py --data

213 Dec 26, 2022

RepMLP: Re-parameterizing Convolutions into Fully-connected Layers for Image Recognition

RepMLP RepMLP: Re-parameterizing Convolutions into Fully-connected Layers for Image Recognition Released the code of RepMLP together with an example o

260 Jan 03, 2023

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

Image Thresholding is used for image segmentation. From a grayscale image, thresholding can be used to create binary images. In thresholding we pick a threshold T.

3 Feb 15, 2022

PyTorch Re-Implementation of EAST: An Efficient and Accurate Scene Text Detector

Description This is a PyTorch Re-Implementation of EAST: An Efficient and Accurate Scene Text Detector. Only RBOX part is implemented. Using dice loss

365 Dec 20, 2022

This repository summarized computer vision theories.

3 Feb 04, 2022

This is used to convert a string to an Image with Handwritten Characters.

Text-to-Handwriting-using-python This is used to convert a string to an Image with Handwritten Characters. text_to_handwriting(string: str, save_to: s

3 Aug 15, 2022

Basic functions manipulating images using the OpenCV library

OpenCV Basic functions manipulating images using the OpenCV library. Reading Ima

3 Feb 17, 2022

This is a Computer vision package that makes its easy to run Image processing and AI functions. At the core it uses OpenCV and Mediapipe libraries.

CVZone This is a Computer vision package that makes its easy to run Image processing and AI functions. At the core it uses OpenCV and Mediapipe librar

648 Dec 30, 2022

Localization of thoracic abnormalities model based on VinBigData (top 1%)

Related tags

Overview

Details

Only inference

Train

Comments

Question to sergey part, 640 or 1024 size?

Regarding Learning Rate

No hyper-parameter yaml found

No MegaMix 341_healthy file

Any pre trained weights