Multi-task Learning "Hydranets" for Autonomous Driving

This project demonstrates the use of a multi-task deep learning algorithm to learn and perform the tasks of semantic segmentation and depth estimation simultaneously.

Project Structure

• cmaps storing cmaps for both the datasets
• models storing the pre-trained models for the hydranets
• notebooks contains the inference and training notebooks
• output contains the output videos and point clouds
• lib contains code for loading the dataset
• lib/network contains the scripts for network architecture
• lib/utils contains the scripts for utility functions

Output (Work in progress)

The current segmentation output is noisy.

References

This project is based on the paper "Real-Time Joint Semantic Segmentation and Depth Estimation Using Asymmetric Annotations". Some of the code has been adapted from the official repository.