Description

Nearly half of the world depends on seafood for their main source of protein. In the Western and Central Pacific, where 60% of the world’s tuna is caught, illegal, unreported, and unregulated fishing practices are threatening marine ecosystems, global seafood supplies and local livelihoods. The Nature Conservancy is working with local, regional and global partners to preserve this fishery for the future.

Currently, the Conservancy is looking to the future by using cameras to dramatically scale the monitoring of fishing activities to fill critical science and compliance monitoring data gaps. Although these electronic monitoring systems work well and are ready for wider deployment, the amount of raw data produced is cumbersome and expensive to process manually.

The Conservancy is inviting the Kaggle community to develop algorithms to automatically detect and classify species of tunas, sharks and more that fishing boats catch, which will accelerate the video review process. Faster review and more reliable data will enable countries to reallocate human capital to management and enforcement activities which will have a positive impact on conservation and our planet.

Machine learning has the ability to transform what we know about our oceans and how we manage them. You can be part of the solution.

Evaluation:

Submissions are evaluated using the multi-class logarithmic loss. Each image has been labeled with one true class. For each image, you must submit a set of predicted probabilities (one for every image). The formula is then,

logloss=−1N∑i=1N∑j=1Myijlog(pij),$$logloss=-\frac{1}{N}\sum _{i=1}^N\sum _{j=1}^M{y}_{ij}\log (p_{ij}),$$

where N is the number of images in the test set, M is the number of image class labels,  log$log$ is the natural logarithm, yij$y_{ij}$ is 1 if observation i$i$ belongs to class j$j$ and 0 otherwise, and pij$p_{ij}$ is the predicted probability that observation i$i$ belongs to class j$j$.

The submitted probabilities for a given image are not required to sum to one because they are rescaled prior to being scored (each row is divided by the row sum). In order to avoid the extremes of the log function, predicted probabilities are replaced with max(min(p,1−10−15),10−15)$max(min(p,1-{10}^{-15}),{10}^{-15})$.

Submission File

You must submit a csv file with the image file name, and a probability for each class.

The 8 classes to predict are: 'ALB', 'BET', 'DOL', 'LAG', 'NoF', 'OTHER', 'SHARK','YFT'

The order of the rows does not matter. The file must have a header and should look like the following:

image,ALB,BET,DOL,LAG,NoF,OTHER,SHARK,YFT
img_00001.jpg,1,0,0,0,0,...,0
img_00002.jpg,0.3,0.1,0.6,0,...,0
...