# BYU - Locating Bacterial Flagellar Motors (V2) — Data

来源：https://www.kaggle.com/competitions/byu-locating-bacterial-flagellar-motors-v2/data

采集时间：2026-10-19T08:01:29Z

As seen below, a diverse assortment of bacterial species showcase the presence of flagellar motors. Our dataset encompasses a few different species. However, for this competition, you can anticipate encountering a predominant number of images featuring the **pseudomonas aeruginosa** species. While the understanding function of flagellar motors is not essential to this contest, it is important to know that flagellar motors generally share a structural core across most species, each species has also evolved variations of the flagellar motor to respond its unique needs. This variety can make visual recognition of the flagellar motor more challenging across species.

![Chen S, Beeby M, Murphy GE, Leadbetter JR, Hendrixson DR, Briegel A, Li Z, Shi J, Tocheva EI, Müller A, Dobro MJ, Jensen GJ. Structural diversity of bacterial flagellar motors. EMBO J. 2011 Jun 14;30(14):2972-81. doi: 10.1038/emboj.2011.186. PMID: 21673657; PMCID: PMC3160247.](assets/6ce85095070846461e94e9a7.jpg)
<small>Chen S, Beeby M, Murphy GE, Leadbetter JR, Hendrixson DR, Briegel A, Li Z, Shi J, Tocheva EI, Müller A, Dobro MJ, Jensen GJ. Structural diversity of bacterial flagellar motors. EMBO J. 2011 Jun 14;30(14):2972-81. doi: 10.1038/emboj.2011.186. PMID: 21673657; PMCID: PMC3160247.</small>

## What should I expect the data format to be?
The **train** directory contains of images of bacteria, with corresponding `x` and `y` points indicating the location of each identified flagella motor. **NOTE: (0,0) is the index of the top left most pixel.** There are also many 'negative' images that do not contain a flagellar motor. Each image was created by averaging 15 tomogram slices.

The train.csv data simplifies the information, where the `image_name` corresponds to the filename of a given image. Each row in train_labels.csv represents a point associated with a flagella motor, and an indication of the presence/absence of a motor denoted by the `visiblemotor` element as a 0 or 1, where 0 indicates no motor and 1 indicates that there is a motor.

The **val** directory contains images of bacteria that are used to score your submission. You should submit a .csv file in the same format indicated below. This .csv file will be scored based off of the labels of these validation images.

## What am I predicting?
You are tasked with predicting points where the flagella motor is located when they are present. If there are no flagella motors in an image, your prediction should include no point (This will be represented in your submission file by a 0 in the motor_visible column and -1.0 in the `x` and `y` columns)

## Files
train.csv: Training data (labels)
train: Directory with 2770 512x512 .png training images of bacteria
sample_submission.csv: Sample submission file in the correct format
validation_labels.csv: You will not have access to this file. This is the file that is used to score your submission. You will see your score based on 50% of the rows in this file. The other 50% of the data will be kept private for use in our final scoring.
validation: Directory containing images that will be used for scoring

## Column info
image_name: Unique image ID
x: the x direction pixel number of motor
y: the y direction pixel number of motor
motor_visible: Contains either 0 or 1. 0 indicates no motor, 1 indicates a motor.

Feel free to explore and utilize the provided files to develop your machine learning models.

Use your Kaggle API key and the command below to download the dataset to your local computer.

NOTE: all images are 512x512