# CRDSM

Crowdsourced Map Land Cover Prediction

167
8
8
12

🛠 Contribute: Found a typo? Or any other change in the description that you would like to see? Please consider sending us a pull request in the public repo of the challenge here.

## 🕵️ Introduction

With tons of satellite images from Open Street Map and only a pc to work with , your mission, should you choose to accept is to identify whether a piece of land is a desert, forest or a water body.
Time to train a classifier with the features extracted and presented in a nice tabular form.

Understand with code! Here is getting started code for you.😄

## 💾 Dataset

• This dataset is crowdsourced from Open Street Map.
• This dataset is derived from geospatial data from two sources:
• Time-series of images captured by landsat satellites from the year 2014 to 2015.
• crowdsourced georeferenced polygons with land cover labels obtained from Open Street Map.
• The crowdsourced polygons cover only a small part of the image area.
• The main challenge with the dataset is that both the imagery and the crowdsourced data contain noise , this is due to cloud cover in the images and inaccurate labeling/digitizing of polygons.
• Each row has `29` attributes.
• `27` attributes describe the time series of NDVI values extracted from the satellite images acquired between `January 2014 and July 2015` in reverse chronological order.
• Dates are given in the format `yyyymmdd`.
• Out of the 2 remaining attributes, one attribute denotes the `Maximum NDVI` (normalized difference vegetation index) value of the corresponding 27 given attributes.
• The last attribute gives the `class` of the land cover in the image. It may be of the following six types: ( forest-0,farm-1,impervious-2, grass-3, water-4,orchard-5)

## Files

Following files can be found in `resources` section:

• `train.csv` - (`10545` samples)This csv file contains the attributes describing the land cover along with the class the land cover belongs to .
• `test.csv` - (`300` samples)File that will be used for actual evaluation for the leaderboard score.

## 🚀 Submission

• Prepare a CSV containing header as `class` and predicted value as class `0`(forest)/`1`(farm)/`2`(impervious)/`3`(grass)/`4`(water)/`5`(orchard) denoting the land cover.
• Name of the above file should be `submission.csv`.
• Sample submission format available at `sample_submission.csv` in the resorces section.

Make your first submission here 🚀 !!

## 🖊 Evaluation Criteria

During evaluation F1 score will be used to test the efficiency of the model where,

$F1 = 2 * \frac{precision*recall}{precision+recall}$

## 📚 References

• Johnson, B. A., & Iizuka, K. (2016). Integrating OpenStreetMap crowdsourced data and Landsat time-series imagery for rapid land use/land cover (LULC) mapping: Case study of the Laguna de Bay area of the Philippines. Applied Geography, 67, 140-149.
• https://latex.codecogs.com/
• https://www.maketecheasier.com/what-is-openstreetmap/