Shypd.aiThe largest open index of AI toolsMonoDepth-PyTorch is an open-source deep learning tool that provides an unofficial PyTorch implementation of the MonoDepth neural network for unsupervised monocular depth estimation. It offers a lightweight model with improved accuracy using ResNet encoders.
Trending
About
MonoDepth-PyTorch is an unofficial PyTorch implementation of the MonoDepth neural network, designed for unsupervised monocular depth estimation. Inspired by the original work of Clément Godard, Oisin Mac Aodha, and Gabriel J. Brostow, this repository aims to provide a more lightweight model with enhanced accuracy. It utilizes ResNet50 and ResNet18 as encoders, with slight modifications and the inclusion of batch normalization for training stability. The tool offers a flexible feature extractor compatible with various torchvision ResNet versions, including options for using pretrained models. It requires stereo-pair images for training and single images for testing, with the KITTI dataset being the primary training data source. The repository includes detailed instructions for data loading, training, and testing, making it accessible for developers and researchers in computer vision.
Capabilities
Pricing & Plans
Open Source
Free
FAQs
Trending
