Help visually impaired navigate their way using stereo images and depth estimation, using triangulation and object detectoin and localization, using state of the art YOLOv5 models.
Stereo depth estimation on the cones images from the Middlebury dataset (https://vision.middlebury.edu/stereo/data/scenes2003/)
Import and use in google colab to avoid any dependancy issuses and enable gpu for faster inference. Run the following in shell to execute code.
python3 obj_det_depth.py
- Hitnet model: https://github.com/google-research/google-research/tree/master/hitnet
- DrivingStereo dataset: https://drivingstereo-dataset.github.io/
- Original paper: https://arxiv.org/abs/2007.12140