Sélectionner une page

EPFL courses

Abstract

This dataset contains the course enrolments of all EPFL students since 2001, and all related information publicly available on the EPFL website, such as the description of the courses they’ve enrolled in

Data Availability

The data still needs to be deposited somewhere.

Example of Challenge

(i) discover relationships between courses, based on student choices and course descriptions
(ii) build a course recommendation system for EPFL students
(iii) build an estimation of course dependencies based on a NLP (Natural Language Processing) analysis of the course descriptions.

Contact

Patrick Jermann and Francisco Pinto

Aerial drone pictures of Savannah in Namibia

Abstract

Near real-time ultrahigh-resolution imaging from unmanned aerial vehicles for sustainable land use management and biodiversity conservation in semi-arid savanna under regional and global change (SAVMAP)

Data Availability

SAVMAP sample available on Zenodo

Example of Challenges

Automatic animal detection in pictures

Contact

Stéphane Joost

ACE wave dataset

Abstract

A large amount of measurements have been gathered with the WaMoS II wave radar during the Antarctic Circumnavigation Expedition (ACE). The WaMoS is an instrument that allows the reconstruction waves by using the backscatter data from the sea surface and directly provides the informations traditionally required by met-ocean engineers (wave spectrum and associated parameters). Unfortunately, the system does not perform correctly in the ice covered ocean due to the backscatter of the ice itself. Under these circumstances the wave characteristics can be inferred by the ship response (heave, roll, pitch, yaw) as measured by the ship GPS.

Data Availability

The data still needs to be deposited somewhere. It will be a subset of the whole expedition.

Example of Challenges

Reconstruct wave conditions from the ship motion during ACE

(i) Identify the transfer function that relates the ship response to the wave conditions using GPS and WaMoS data in open water

(ii) Apply the transfer function to reconstruct the wave climate in the ice covered ocean (where only GPS information are available).

Contact

Alberto Albarello

ACE fish dataset

Abstract

A unique data set of a proxy for biomass was collected during the Antarctic Circumnavigation Expedition (ACE), along with CTD data and satellite remote sensing. We also collected position, depth and accelerometry data from predators to better understand predator-prey interactions and inform fisheries and conservation management.

 

Data Availability

Stay tuned! The data will be deposited in the coming days. Will be a subset of the whole expedition comprising all days of EK80 200 kHz echo-sounder data during the leg 2 of ACE (from Australia to South America).

Example of Challenges

(i) Create a method that is adapted to detect DISCRETE krill swarms in order to extract their key characteristics. For all swarms along the ship track, we want latitude, longitude and time stamp (from the ship data) but also duration, height, length, other morphometric indicators if possible, mean depth and intensity.
(ii) Identify the environmental predictors of the krill swarm characteristics using CTD data and satellite images. During the voyage, the temperature (°C), the salinity (PSU), the dissolved oxygen, and the fluorescence profiles were measured.

Contact

Camille Le Guen

Wifi Traces from EPFL Campus

Abstract

The dataset contains Wifi traces, a pedestrian routing graph and potential attractivity measures of EPFL campus. The data have been used in Danalet et al., 2014 (link: http://dx.doi.org/10.1016/j.trc.2014.03.015), where stops and activities at these stops have been detected. These stops and activities have been used in Danalet et al., 2016 (link: https://dx.doi.org/10.1016/j.jocm.2016.04.003), where a location choice model has been developed for EPFL catering locations.

Data Availability

The raw data are available here (link: http://doi.org/10.5281/zenodo.15798) and the location choice model is available here (link: http://doi.org/10.5281/zenodo.1038622).

Example of Challenge

(i) Detect group of people going together to restaurant for lunch break or to Sat’ for an afterwork beer.
(ii) Add extra factors in the catering location choice model (on top of beer availability), such as an estimation of queuing time.

Contact

Antonin Danalet

FMA: A Dataset For Music Analysis

Abstract

We introduce the Free Music Archive (FMA), an open and easily accessible dataset which can be used to evaluate several tasks in music information retrieval (MIR), a field concerned with browsing, searching, and organizing large music collections. The data is made of 106,574 tracks, 16,341 artists, 14,854 albums, arranged in a hierarchical taxonomy of 161 genres, for a total of 343 days of audio and 917 GiB, all under permissive Creative Commons licenses.

Data Availability

Code, data, and usage examples are available at here.

Example of Challenge

(i) given the audio (with metadata or not), predict the genre, artist, year, and/or tags. Genre is especially well suited as the dataset features a hierarchical taxonomy of 161 genres.
(ii) visualization: embed (using the audio and/or the metadata) and visualize the 100k tracks in a 2D/3D space. What kind of structure arise? Do we find clusters of genres or tags, or something else entirely?
 Many more ideas can be explored, including those mentioned in the “Usage” section of the original paper: https://arxiv.org/abs/1612.01840

Contact

Michael Defferrard