taarun-srinivas / semantic-segmentation-for-visual-place-recognition-vpr-

generate semantic data in different seasonal modalities for an simulated autonomous car

Semantic-Segmentation-for-Visual-Place-Recognition-VPR-

Idea

The main idea of this work is to generate semantic data for training an autonomous car, detect and have a good understanding of it's environment. In this work, I have generated semantic data for different seasonal modalities. These seasonal modalities are clear night, clear sunset, rain noon, and wet cloudy weather. This algorithm is built on top of STEGO algorithm. The objective of this work is to use this semantic data to train or evaluate Visual Place Recognition (VPR) models.

requirements

• Python - 3.9
• OpenCV - 4.6.0
• Pytorch - 1.13.1 + cu116

Visualization

• Clear Night

• Clear Sunset

• Rain Noon

Acknowledgement

I wish to extend my utmost gratitude for the invaluable support and guidance offered by the AI4CE lab at NYU Tandon. I would like to express special appreciation to Professor Chen Feng for his unwavering support and insightful advice throughout the course of this project. My sincere thanks go to Mr. Chao Chen for mentoring and providing guidance in the successful completion of this endeavor.