Ishana team ML_enthusiasts

A winning solution for the Takshak '21 Ishana (Computer Vision Challenge) organized by RoboISM, IIT(ISM) Dhanbad.

Team members : Bhavya Dalal, Tanishq Selot, Kshitij M Bhat

Link to Presentation : Presentation.pdf

Requirements

• Python 3.7.12
• Numpy
• Pytorch 0.2.0
• OpenCV
• Cuda 8.0
• NetworkX
• Scikit-Learn

Download the trained model

D-LinkNet34 model

Follow the Main.ipynb in Google Colaboratory for using the model.

References

• DeepGlobe Road Extraction Challenge
• L. Zhou, C. Zhang, and M. Wu, “D-linknet: Linknet with pretrained encoder and dilated convolution for high resolution satellite imagery road extraction,” in Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR) Workshops, June 2018. Open Access Link Code