Bosch-ELTE Student Competition for Parking Car Detection
The result of the first Bosch-ELTE competition is as follows:
1) Barbara Szabó
2) Dávid Fischer
3) Bence Csabai
Congratulations for the students.
The original proposal can be read here:
It is our pleasure to announce the first Bosch-ELTE competition for parking car detection. 😊
The aim of this project is to detect parking cars in the given 3D scenes. Camera images and/or 3D point clouds, scanned by a LiDAR device, can be applied. Three cameras and a LiDAR are mounted on the top of ELTECar they serve the input data for the detection.
Preliminary data can be downloaded from here: http://cv.inf.elte.hu:8000/data_sample
As you can see from the published data, there are six different places and two complete scans at each location. One scan is taken at many different timestamps. But the bounding boxes must be determined only at the last timestamp of the short sequence.
LiDAR point clouds are given in text files. Each row is an individual 3D point, a point has six coordinates: X, Y, Z, intensity, intensity, intensity. The last three coordinates are the same, they are the LIDAR reflectance at the measured point. X, Y, Z, are the 3D point coordinates. The origin is the center of the LiDAR device. Remark that the LiDAR often produces [0,0,0] coordinates, that should be filtered out.
The parking cars should be represented by four corners of their bounding boxes on the ground. They can be given either in LiDAR 3D coordinates; or four locations in any of the camera images, given by pixel coordinates.
This competition is open for every BSc and MSc students of the Faculty of Informatics at the Eötvös Loránd University.
Deadline of the implementation is
22nd April. extended to 1st May.
At the deadline, participants get another six scans that are recorded at the same locations but at different time. They should detect the bounding boxes and submit the results within 30 minutes.
The results are evaluated by our tool. It is prohibited to use manual tools for computing the results, the whole parking detection process should be fully automated. Students with the best results should personally present their solutions again to the organizers, in real time.
The competition is sponsored by the Robert Bosch GmbH. The first, second, and third competitors are awarded:
1st prize: Trecam charger bluetooth speaker + Bloofi bluetooth headphones + Temboo thermometer vacuum flask
2nd prize: Briden backpack
3rd prize: Dickens power bank
The schedule of the competition is as follows:
- Data available on 25th February.
- Calibration data for the setup available on 4th March. Document for the sensor kit is also available.
- Scoring system is also available, it is published on 11th March.
- Annotated bounding boxes are published on 18th March. File format description can also be downloaded.
- Submission deadline is
22nd April.!!! Extended to 1st May
- The conditions, file formats, content, etc, for the final submission is announced at the end of April.
- The final testing data are available at our website. Click here to download. The detection results, that are short text files including the detected bounding box corners, should be submitted until
17:00PM21:00 PM by email to email@example.com (Subject: [Bosch-ELTE Parking cars] please). Instructions for the submission format are also available now.
- Deadline is over, we have received three submissions. Results ceremony is going to be on 9th May.
If you want to get the latest information about the competition, join our channel in Teams.
Organizer: Geometric Computer Vision Group
Contributors: Tekla Tóth, Viktória Szekeres, Bandó Kovács, Tamás Tófalvi, Dominik Molnár, Áron Varga, Tamás Fischl, Péter Pozsegovics, Levente Hajder, János Szalai-Gindl, Dániel Varga