Self-driving cars collect a large amount of visual data for decision-making. It uses three tools to replicate human eyes - cameras, radar, and lidar. Together they create a complete picture of the car’s environment. They help the car identify the place, speed, and objects around it.
Understanding the environment around the self-driving car is really critical to driving safely on the roads. Having an understanding of individual elements such as sidepaths, buildings, lanes, trees, people and other objects around the road allows safe driving. With detailed segmentation of the scenes, self-driving cars will become more accurate & safer.
For this puzzle, your task is to segment the scene of the input image. The starter kit gives a walkthrough of image segmentation and its basics.
✔ The Task
In this challenge your task will be to build an automated algorithm that will take the road scene image and will output the semantic segmentation from the input image :
In machine learning terms: this is a multi-class semantic segmentation task.
🚀 Getting Started
Make your first submission using starter kit. 🚀
The dataset files are available here.
The images ( RGB format ) are in image folder where the corresponding semantic segmentation labels ( grayscale format ) are in the segmentation folder.
data ├── train.zip │ ├── image │ │ ├── 0.jpg │ │ ├── 1.jpg │ │ ├── ... │ │ └── 3999.jpg │ └── segmentation │ ├── 0.png │ ├── 1.png │ ├── ... │ └── 3999.png └── test.zip └── image ├── 0.jpg ├── 1.jpg ├── ... └── 999.jpg
Following files are available in the
train.npz- ( 4k samples ) The training images and corresponding semantic segmentation labels.
test.zip- ( 1k samples ) It contains the images for testing data and the generate labels will be used in the leaderboard.
📨 How to submit
Make your first submission using the starter kit 🚀
segmentationfolder and fill the corresponding segmentation image.
Inside a submission directory, put the
.ipynbnotebook from which you trained the model and made inference and save it as
Zip the submission directory
Overall, this is what your submission directory should look like -
submission.zip ├── segmentation │ ├── 0.png │ ├── 1.png │ ├── ... │ └── 999.png └── notebook.ipynb
🖊 Evaluation Criteria
The evaluation metric for this competition is F1 Score with
[Getting Started Notebook] Semantic Segmentation