ML-CDS 2019: Challenge

Challenge Submissions

To be evaluated, the participating teams will submit their model together with their inference code in their preferred framework from the supported frameworks in this challenge (Keras, Tensorflow, and PyTorch). Please include a 'requirements.txt' file with your submission. This file will be used to build the virtual environment that will run your code. In addition, each participating team will submit 1-2 paragraphs with a short explanation on the proposed solution and a few lines of instructions how to run the code.

All of the files, including the model, needs to be zipped and uploaded to the private box folder that will be assigned to each team, where the zip folder should contain the team name, for example 'TheBestTeam_submission.zip'. The link to the box folder will be sent to the team's contact person after registering to the challenge. Every submission will be accessible only by the challenge organizers, and will not be shared with other teams.

The inference code will take care of loading the model, doing all the preprocessing needed, running prediction on the image, and saving the results. The input to the inference code will be a text file with a list of all the images to run prediction on, whereas the output will be a CSV file with the first column for the image names followed by columns per label. The predictions in the CSV files should be saved as probabilities, sigmoid values, or other types of output values, based on the proposed solution.

To standardize the submissions, the inference code should be denoted 'test.py', where the syntax to run the inference code should be as follows:

python-test-path-to-image-list

For solutions intended to run on a GPU, please include an additional input variable to specify the GPU id (0, 1, 2, etc) to execute the code on:

python-test-path-to-image-list-gpu

Please specify in your documentation that your code has GPU support.