The benefits of forests are well-known to all. It sustains a healthy ecosystem and maintains the earth’s temperature. In the recent decade, large scale deforestation has caused massive displacement of wildlife and an increase in global temperature.
In large forests, it is often hard to predict the density of trees and the distribution of various species. This information is necessary to maintain balance in the ecosystem. Through computer vision, we are trying to address this challenge. Given aerial images of a forest, can you segment various trees into groups using deep learning and computer vision?
Here’s the starter kit to give you a heads up!
If you are a beginner, check out our blog on Computer Vision which has a similar problem from Blitz 3 that maps water bodies using satellite images.
💪 Getting Started
These images are captured by drone in the forest and you will need to predict the segmentation of a specific kind of tree in the forest.
Use our Getting Started Notebook available here.
In this dataset, for a given data file, such as train.zip. It will contain two folders, images & segmentations. And inside both folders, the image name in the images folder will be corresponding to the name of segmentation images in the segmentations folder. Like below -
Things to make sure -
- Segmentation images are in .png extension
- Segmentation images are in grayscale and only contain pixel values 0 and 255.
Following files are available in the resources section:
- train.zip - [ 5000 samples ] Used for Training. Include labels.
- test.zip - [ 5000 samples ] Used for evaluation. Does not include labels.
- Creating a submission directory
- Use test.zip and predict the corresponding labels in a segmentation folder.
- Put the segmentation folder in the submission directory.
- Inside a submission directory, put the .ipynb notebook from which you trained the model and made inference and save it as original_notebook.ipynb.
- Overall, this is what your submission directory should look like
- Zip the submission directory!
Make your first submission here 🚀 !!
🖊 Evaluation Criteria
During the evaluation, Dice Coefficient over-segmentation images will be used to test the efficiency of the model.
- 💪 Challenge Page: https://www.aicrowd.com/challenges/trees-segmentation
- 🗣️ Discussion Forum: https://www.aicrowd.com/challenges/trees-segmentation/discussion
- 🏆 Leaderboard: https://www.aicrowd.com/challenges/trees-segmentation/leaderboards