Skip to content

Submission Details

Jump to bottom
bbischke edited this page Jun 6, 2017 · 4 revisions

DIRSM Submission

Run Submissions

For the DIRSM subtask, you may submit up to 5 runs. Upload your runs (please see the format below) to your team folder. Please be sure to submit the following types of runs:

  1. Required run: automated using visual information only;
  2. Required run: automated using metadata information only;
  3. Required run: automated using metadata-visual fused without other resources than provided by the organizers;
  4. General run: everything automated allowed, including using data from external sources (e.g. Twitter, Flickr, Wikipedia);
  5. General run: everything automated allowed, including using data from external sources (e.g. Twitter, Flickr, Wikipedia);

Please note that for generating runs 1 to 3 participants are allowed to use only information that can be extracted from the provided data (e.g., provided content descriptors, content descriptors of their own, etc). Text and visual descriptors provided with the data are intended to help all teams, regardless their level of expertise in the area of text or image processing. For runs 4 & 5 everything is allowed, both from the method point of view and the information sources.

In the event that a team is unable to submit one of the required runs, please contact the organizers.

Submission Format

Please submit your runs for the SMIR task in the form of a text file, where each line contains one image_id. Please make sure you order the image_ids descending by their relevance of showing evidence of a flooding event. An example should look like this (where the image_id in the first line corresponds to an image with strong direct evidence of flooding and the one in the last line to a non flooding):

238473289 
726183944
…
193846823

Please note that no white spaces or other special characters are allowed. When you create the filenames for your runs, please follow this pattern:

ME17MST_DIRSM_YYY_ZZZ.txt

where ME17MST is the code of the task, DIRSM is the name of the subtask, YYY is the acronym of your team and ZZZ is the name of your run.

FDSI Submission

Run Submissions

For the FDSI subtask, you may submit up to 5 runs. Upload your runs (please see the format below) to your team folder. Please be sure to submit the following types of runs:

  1. Required run: automated using satellite information only;
  2. Required run: automated using satellite information only;
  3. Required run: automated using satellite information only;
  4. General run: everything automated allowed, including using data from external sources (e.g. Open Street Map, Elevation Maps, additional Satellite Images);
  5. General run: everything automated allowed, including using data from external sources (e.g. Open Street Map, Elevation Maps, additional Satellite Images);

Please note that for generating runs 1 to 3 participants are allowed to use only information that can be extracted from the provided data. For runs 4 & 5 everything is allowed, both from the method point of view and the information sources.

In the event that a team is unable to submit one of the required runs, please contact the organizers.

Submission Format

Please submit your runs for the SMIR task in the form of a zip file. When you create the filenames for your runs, please follow this pattern:

ME17MST_FDSI_YYY_ZZZ.zip

where ME17MST is the code of the task, FDSI is the name of the subtask, YYY is the acronym of your team and ZZZ is the name of your run. The zip file should contain one folder for each flooding event with one segmentation mask for each image patch.

The submitted segmentation mask should follow the same data format as the provided ground truth mask (shape 1 x height x width where each pixel is an integer label for the class, 0=background, 1=flooded area). Please name the segmentation mask with the following pattern:

seg_mask_ZZZ.png

where ZZZ corresponds to the image_id (filename without extension) of each image patch.

Clone this wiki locally