Competition: Can I make a wish? Predicting the presence of meteors in images
BRACIS 2017 - Brazilian Conference on Intelligent Systems
SBBD 2017 - Brazilian Symposium on Databases
KDMiLe 2017 - 5th Symposium on Knowledge Discovery, Mining and Learning
October 2nd-5th, 2017, Uberlândia, MG, Brazil
This competition is part of the 1st KDD-BR (Brazilian Knowledge Discovery in Databases) competition, one of the joint activities of the 2017 editions of the Brazilian Conference on Intelligent Systems (BRACIS), the Brazilian Symposium on Databases (SBBD) and the Symposium on Knowledge Discovery, Mining and Learning (KDMiLe).
This competition involves classifying images captured by a monitoring station installed at the UNIVAP (University of Vale do Paraíba) astronomy observatory, at São José dos Campos, São Paulo, Brazil. This station composes one of the monitoring stations of the EXOSS Citizen Science organization. This Brazilian non-profit organization aims to monitor meteors (popularly kown as shooting stars) which cross the southern skies. It is a network currently composed of 50 active monitoring stations at various locations of the Brazilian territory. Any interested citizen can apply as a potential contributor.
The data acquisition is made by a simple and low cost video monitoring camera, usually used in security moniotring systems, with a dedicated software. Each time a moving object is detected by the camera, it starts shooting. This includes meteors (Figure 1a), but also birds, insects, planes (Figure 1b) and even lightning thunders. Some days are cloudy, others rainy, there are varying stars in the sky, and there are many other challenging situations. The camera also takes some noisy shootings sometimes. The system records a video and also an image which is a composition of the video frames, which will be used at the competition. Since the interest is to monitor the occurence of meteors, this competition aims to built an automate system for classifying the recorded images into two classes: meteor vs non-meteor.
The images were captured in 2017 during the months of April and May and were categorized by Jennifer Nielsen, an Aeronautics Engineering student at UNIVAP, under the supervision of Dr Irapuan Rodrigues, Physics and Astronomy Professor at UNIVAP. The "!UFO Capture" software was used to capture the videos and images and the "UFO Analyzer" software was used for cleaning and analyzing the trajectories. Both softwares are part of the Sonotaco package.
The submissions will be closed at September 18th, 2017. Therefore, the final results will be based on the solutions posted until September 18th, 2017. Please note that the system will still be able to receive submissions after this deadline, but they will be not taken into account for computing the final results.
As this competition is part of the joint activities of the 2017 editions of the Brazilian Conference on Intelligent Systems (BRACIS), the Brazilian Symposium on Databases (SBBD) and the Symposium on Knowledge Discovery, Mining and Learning (KDMiLe), the top three teams will be invited to present their solutions at a competition award session, at October 2nd, 2017. The actual final positions of the ranking will be disclosed during the conference dinner.
We are glad to anounce the three top teams of the 1st KDD-BR competition. The session dedicated to the competition will be in October, 3rd, from 6pm to 7pm - Room: B1 - Center Convention.
The winning teams, in alphabetical order, are:
Artificial Psycho Killers - Humberto Brandão and Hugo Pinto
ML-UFSCAR - Renato Moraes Silva, Tiago Almeida and JLochter
Victor Almeida - UFF and Petrobras
The program for the session is:
18h00: Competition presentation
18h15: Team Artificial Psycho Killers
18h30: Team ML-UFSCAR
18h45: Victor Almeida - UFF and Petrobras
Each team will have a 15 minute time slot. We recommend 10 minutes of presentation and 5 minutes for questions.
The final positions of the teams will be disclosed at the conference dinner, at October, 4th.