Repository Issue Activity (beta)

#168 Is MapAnything depth map affine invariant?

#166 Support pycolmap 4.x for COLMAP export

#164 Support for ONNX / TensorRT conversion

#149 Forced weight download for dinov2 from torch hub

#8 The optimal image aspect ratio for single-image calibration.

#33 Inputting internal and external parameters images, the output does not correspond to the GT lidar data

#40 Trajectory Drift and Point Cloud Distortion on KITTI 06 During Sharp Turns

#49 Support for rigid camera rig as input?

#61 purpose of recover_pinhole_intrinsics_from_ray_directions

#74 Prediction with sparse depth is worse than without it

#81 Metric scale is underestimated by ~3

#86 How to set foreground mask for input

#88 Nice speedups possible with Block-Sparse Attention

#90 Request for Guidance on Using a Custom Dataset with the WAI Benchmarking Framework

#93 Reconstruction Scale/Pose Mismatch: Predicted Poses and Intrinsics Not Aligning with Ground Truth

#98 Bad output point cloud when passing depth as input.

#107 Question about obtaining correctly scaled point clouds

#110 tips for resampling/balancing across datasets when mixing training data

#111 Why make vggt and pi3 more powerful using mapanything framework

#124 Very large loss of "FactoredGeometryScaleRegr3DPlusNormalGMLoss_pts3d_conf_loss" when fine tuning

#126 how can i mitgate this issue when scanning area ( walls not overalping )

#127 Sliding window chaining of predicted camera poses

#134 Fine-tuning custom datasets with noisy pose/depth inputs: what's best for noise mitigation?

#141 External predictions with `mapanything` model not working

#147 Nice point maps. Yet, deceiving point map reprojection consistency.