Pretrained model for "Toward Practical Monocular Indoor Depth Estimation" (DistDepth, CVPR 2022)

Pretrained models for "Toward Practical Monocular Indoor Depth Estimation" (DistDepth, CVPR 2022) by Cho-Ying Wu, Jialiang Wang, Michael Hall, Ulrich Neumann, Shuochen Su.

• Official implementation and pre-trained models: https://github.com/facebookresearch/DistDepth
• Small changes to export via torch script tracing.
• The exported model currently fails with GPU inference, so only CPU inference is enabled

Pretrained model for "Revisiting Stereo Depth Estimation From a Sequence-to-Sequence Perspective with Transformers" (ICCV 2021)

Pretrained models for "Revisiting Stereo Depth Estimation From a Sequence-to-Sequence Perspective with Transformers" (ICCV 2021) by Li, Zhaoshuo and Liu, Xingtong and Drenkow, Nathan and Ding, Andy and Creighton, Francis X. and Taylor, Russell H. and Unberath, Mathias.

• Official implementation and pre-trained models: https://github.com/mli0603/stereo-transformer
• Made some small changes to allow torch script export via tracing.
• The exported model currently fails with GPU inference, so only CPU inference is enabled (much slower for this method)

Pretrained model for "RAFT-Stereo: Multilevel Recurrent Field Transforms for Stereo Matching"

Pretrained models for "RAFT-Stereo: Multilevel Recurrent Field Transforms for Stereo Matching." (3DV 2021) by Lahav Lipson, Zachary Teed and Jia Deng.

• Official implementation: https://github.com/princeton-vl/RAFT-Stereo
• Pytorch script models generated by tracing from nburrus/RAFT-Stereo@ebbb5a8

Extra notes:

• cpu models all use --corr_implementation reg because alt exploded in memory for some reason during the export.
• cuda models all use --corr_implementation reg too, alt failed when running the exported script
• The fast version does NOT use the optimized cuda sampler mentioned here.
• ONNX export will likely work with pytorch 1.12 and op16, but it failed so far because of the grid sampling.

Pretrained model for "HITNet: Hierarchical Iterative Tile Refinement Network for Real-time Stereo Matching."

Weights for "HITNet: Hierarchical Iterative Tile Refinement Network for Real-time Stereo Matching." (CVPR 2021) by Vladimir Tankovich, Christian Häne, Yinda Zhang, Adarsh Kowdle, Sean Fanello, Sofien Bouaziz.

• Official implementation: https://github.com/google-research/google-research/tree/master/hitnet
• ONNX models from https://github.com/PINTO0309/PINTO_model_zoo/tree/main/142_HITNET
• Inference sample code adapted from https://github.com/ibaiGorordo/ONNX-HITNET-Stereo-Depth-estimation

Pretrained model for "Attention-Aware Feature Aggregation for Real-time Stereo Matching on Edge Devices" (ACCV 2020))

Weights for "Attention-Aware Feature Aggregation for Real-time Stereo Matching on Edge Devices" (ACCV 2020) by by Jia-Ren Chang, Pei-Chun Chang and Yong-Sheng Chen.

Official implementation: https://github.com/JiaRenChang/RealtimeStereo

Traced with pytorch script after slightly adjusting the code (see PR JiaRenChang/RealtimeStereo#15).

CREStereo Pre-trained Models

Weights for Practical Stereo Matching via Cascaded Recurrent Network with Adaptive Correlation (CVPR 2022).

• Official implementation: https://github.com/megvii-research/CREStereo
• Port to ONNX + loading code: https://github.com/ibaiGorordo/ONNX-CREStereo-Depth-Estimation
• Pre-trained model zoo: https://github.com/PINTO0309/PINTO_model_zoo/tree/main/284_CREStereo