https://github.com/innolitics/kaggle-hubmap

HuBMAP - Hacking the Human Vasculature (Kaggle Competition)
https://github.com/innolitics/kaggle-hubmap

Last synced: about 2 months ago
JSON representation

HuBMAP - Hacking the Human Vasculature (Kaggle Competition)

• Host: GitHub
• URL: https://github.com/innolitics/kaggle-hubmap
• Owner: innolitics
• Created: 2023-06-12T15:13:39.000Z (about 3 years ago)
• Default Branch: main
• Last Pushed: 2023-07-05T18:20:47.000Z (almost 3 years ago)
• Last Synced: 2025-06-01T21:47:24.998Z (about 1 year ago)
• Language: Jupyter Notebook
• Size: 9.88 MB
• Stars: 1
• Watchers: 2
• Forks: 1
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

          
# kaggle-HuBMAP

HuBMAP - Hacking the Human Vasculature (Kaggle Competition)

## Kidney microscopy resources

- [kidney under a microscope](https://www.youtube.com/watch?v=m-xdxuF4_lE)

    - glomerulus: "little ball of yarn"

    - distal convoluted tubules: "open centers"

    - proximal convoluted tubules: "fluffy centers"

## Image data overview

- training:

    - metadata summary csv: `metadata/trn_img_meta_7033.csv`

    - 7033 samples

    - 512 x 512 x 3

    - data type `uint8`

    - mean intensity across 3 channels:

    

    

    - stdev intensity across 3 channels:

    

    - sample image:

    

- test:

    - 1 sample

    - 512 x 512 x 3

    - data type `uint8`

    - mean 163.5

    - std 53.1

    - sample image:

     

## Provided CSVs

- tile:

    - location: `metadata/tile_meta_7033.csv`

    - columns:

        - id

        - source_wsi (patients from where WSIs were extracted)

            - 1-4, 6-14

            - source_wsi 1-4 described in `wsi_meta`

        - dataset:

            1. 422

            2. 1211

            3. 5400

    - dataset description:

        1. Annotations that have been expert reviewed.

        2. Comprises the remaining tiles from these same WSIs and contain sparse annotations that have not been expert reviewed.

        3. Tiles extracted from an additional nine WSIs. These tiles have not been annotated. You may wish to apply semi- or self-supervised learning techniques on this data to support your predictions.

        - i, j: upper-left corner within the WSI where the tile was extracted

- wsi (whole slide images):

    - source_wsi (4 patients: 3 females, 1 male)

    - age

    - sex

    - race

    - height (cm)

    - weight (kg)

    - bmi

## Annotations

- Source: `polygons.jsonl`

- 1633 unique IDs

- 3 labels:

    1. blood_vessel

    2. glomerulus

    3. unsure

- produced 1633 segmentations masks with name `mask.tif`

- sample image with segmentation: