Deep Learning against cancer. Intel Contest

Cervical cancer is a terrible disease that kills hundreds of thousands of lives each year. But even worse is that most of these lives could be saved. Cervical cancer develops very slowly, and if a tumor is detected in the first 5 years after its appearance, the chance of survival is almost 100%. Thus, regular examinations can lime this disease in the bud. Unfortunately, 85% of women in the world have regular medical care not available. They can be saved by the technology of machine recognition of photographs, reducing the requirements for the qualifications of medical personnel and increasing the percentage of correct diagnoses. It is these recognition algorithms that are created in the competition of programmers Intel & MobileODT Cervical Cancer Screening , in which we invite you to participate.

The competition is held at the kaggle site in cooperation with MobileODT , the manufacturer of the EVA digital colposcope. Its essence is as follows. It is necessary to find the best algorithm for machine recognition of the shape of the cervix of a particular patient - this is required for making a correct diagnosis. A set of test colposcopic images, already sorted by type, is proposed as the source material for the work.

The total prize fund of the contest is $ 100,000 , of which the winner will receive $ 50,000 . Intel has established a special prize of $ 20,000 for the participant who best used its software. First of all, Intel Deep Learning SDK is a set of tools for developing Deep Learning applications. Further, Intel Optimized Caffe - a framework for convolutional, fully connected and neural networks can be useful to you. Finally, you can develop your own framework using such productivity tools as Intel Math Kernel Library-DNN or Intel Distribution for Python . An overview of the proposed Intel products with links to training resources on them is available on a special project page .
')
In addition, as part of a joint educational project with Colfax Research , which we already talked about , Intel provides contestants with access to a cluster of servers based on Intel's latest Xeon Phi. So you can run your models on the most modern and powerful hardware with Intel performance libraries pre-installed on it.

To start participating in the competition, you need to register at the kaggle site, create a team (or join existing ones, there are already over 500 of them), download the initial data and ... go ahead! Perhaps your developer skill will save more than one life.

The introductory stage will last until June 14 . We (the Intel blog team) will be very pleased if the developers from Russia, who will take part in the competition, decide to write about their experience here - we will provide an opportunity.