Drivable Area Detection using Image Segmentation

The aim of project is to predict drivable area in an image using Unet.

Model

Model can be downloaded from here https://drive.google.com/file/d/1QZbHF8O6NgpbDaN2fDe40h0L1J088pln/view?usp=sharing

Dataset

The dataset can be found in the course at this address. It's in Pickle format and contains 3,000 images and labels downloaded from Berkeley Deep Drive Dataset.Then this dataset is further increased to more 3000 images using data augmentation.

Dataset download Link : [https://drive.google.com/file/d/14CLaT5mipIQlKcNQNlnV5B6hu6ziqYBj/view?usp=sharing]

Test

To test the result please refer ipynb file attached and trained model.

Results

The mask and original image is blended using cv2.bitwise_and technique.

references

Unet: https://arxiv.org/abs/1505.04597

https://courses.thinkautonomous.ai/image-segmentation