{ "data": { "aspect_ratio_range": [0.5, 2.0], "area_range": [250000, 1000000], "clamp_max_depth": 1000.0, "center_augmentation": 0.5, "fov_range_absolute": [1, 179], "fov_range_relative": [0.01, 1.0], "image_augmentation": ["jittering", "jpeg_loss", "blurring"], "datasets": [ { "name": "A2D2", "path": "path/to//A2D2/", "label_type": "C", "weight": 0.8, "depth_unit": 1, "fov_range_relative": [0.5, 1.0], "image_augmentation": ["jittering", "jpeg_loss", "blurring", "dof"], "depth": "depth_completed.png" }, { "name": "ARKitScenes", "path": "path/to//ARKitScenes/", "label_type": "B", "weight": 8.6, "depth_unit": 0.001, "fov_range_relative": [0.5, 1.0], "image_augmentation": ["jittering", "jpeg_loss", "blurring", "dof"], "depth": "depth_completed.png" }, { "name": "Argoverse2", "path": "path/to//Argoverse2/", "label_type": "C", "weight": 7.4, "depth_unit": 1, "fov_range_relative": [0.5, 1.0], "image_augmentation": ["jittering", "jpeg_loss", "blurring", "dof"], "depth": "depth_completed.png" }, { "name": "MegaDepth", "path": "path/to//MegaDepth_840/", "label_type": "C", "weight": 5.6, "fov_range_relative": [0.5, 1.0], "image_augmentation": ["jittering", "jpeg_loss", "blurring", "dof"], "depth": "depth_completed.png" }, { "name": "Taskonomy", "path": "path/to//Taskonomy/", "label_type": "B", "weight": 10.0, "depth_unit": 1, "fov_range_relative": [0.75, 1.0], "image_augmentation": ["jittering", "jpeg_loss", "blurring", "dof"], "finite_depth_mask": "only_known", "depth": "depth_completed.png" }, { "name": "Waymo", "path": "path/to//Waymo/", "label_type": "C", "weight": 6.4, "depth_unit": 1, "fov_range_relative": [0.5, 1.0], "image_augmentation": ["jittering", "jpeg_loss", "blurring", "dof"], "depth": "depth_completed.png" }, { "name": "ScanNetpp", "path": "path/to//ScanNetpp/", "label_type": "B", "weight": 4.8, "depth_unit": 1, "fov_range_relative": [0.33, 1.0], "image_augmentation": ["jittering", "jpeg_loss", "blurring", "dof"], "depth": "depth_completed.png" }, { "name": "BlendedMVS", "path": "path/to//BlendedMVS/", "label_type": "B", "weight": 12.0, "fov_range_relative": [0.5, 1.0], "image_augmentation": ["jittering", "jpeg_loss", "blurring", "dof"], "depth": "depth_completed.png" }, { "name": "ObjaverseV1", "path": "path/to//ObjaverseV1/", "label_type": "A", "weight": 4.8, "center_augmentation": 0.25, "fov_range_relative": [0.7, 1.0], "image_augmentation": ["jittering", "jpeg_loss", "blurring", "shot_noise"] }, { "name": "GTA-SfM", "path": "path/to//GTA-SfM/", "label_type": "A", "weight": 2.8, "depth_unit": 1, "fov_range_relative": [0.5, 1.0], "image_augmentation": ["jittering", "jpeg_loss", "blurring", "shot_noise", "dof"] }, { "name": "Hypersim", "path": "path/to//Hypersim/", "label_type": "A", "weight": 5.0, "depth_unit": 1, "fov_range_relative": [0.5, 1.0], "image_augmentation": ["jittering", "jpeg_loss", "blurring", "shot_noise", "dof"], "finite_depth_mask": "only_known" }, { "name": "IRS", "path": "path/to//IRS/", "label_type": "A", "weight": 5.6, "depth_unit": 1, "fov_range_relative": [0.5, 1.0], "image_augmentation": ["jittering", "jpeg_loss", "blurring", "shot_noise", "dof"], "finite_depth_mask": "only_known" }, { "name": "KenBurns", "path": "path/to//KenBurns/", "label_type": "A", "weight": 1.6, "fov_range_relative": [0.75, 1.0], "image_augmentation": ["jittering", "jpeg_loss", "blurring", "shot_noise", "dof"] }, { "name": "MatrixCity", "path": "path/to//MatrixCity/", "label_type": "A", "depth_unit": 1, "weight": 1.3, "fov_range_relative": [0.33, 1.0], "image_augmentation": ["jittering", "jpeg_loss", "blurring", "shot_noise", "dof"] }, { "name": "MidAir", "path": "path/to//MidAir/", "label_type": "A", "depth_unit": 1, "weight": 4.0, "fov_range_relative": [0.33, 1.0], "image_augmentation": ["jittering", "jpeg_loss", "blurring", "shot_noise", "dof"] }, { "name": "MVS-Synth", "path": "path/to//MVS-Synth/", "label_type": "A", "depth_unit": 0.1, "weight": 1.2, "fov_range_relative": [0.33, 1.0], "image_augmentation": ["jittering", "jpeg_loss", "blurring", "shot_noise", "dof"] }, { "name": "Structured3D", "path": "path/to//Structured3D/", "label_type": "A", "weight": 4.8, "depth_unit": 0.001, "fov_range_relative": [0.5, 1.0], "image_augmentation": ["jittering", "jpeg_loss", "blurring", "shot_noise", "dof"], "finite_depth_mask": "only_known" }, { "name": "Synthia", "path": "path/to//Synthia/", "label_type": "A", "depth_unit": 1, "weight": 1.2, "fov_range_relative": [0.75, 1.0], "image_augmentation": ["jittering", "jpeg_loss", "blurring", "shot_noise", "dof"] }, { "name": "TartanAir", "path": "path/to//TartanAir/", "label_type": "A", "depth_unit": 1.0, "weight": 5.0, "fov_range_relative": [0.5, 1.0], "image_augmentation": ["jittering", "jpeg_loss", "blurring", "shot_noise", "dof"] }, { "name": "UrbanSyn", "path": "path/to//UrbanSyn/", "label_type": "A", "weight": 2.1, "depth_unit": 1, "fov_range_relative": [0.5, 1.0], "image_augmentation": ["jittering", "jpeg_loss", "blurring", "shot_noise", "dof"] }, { "name": "ApolloSynthetic", "path": "path/to//ApolloSynthetic/", "label_type": "A", "weight": 4.0, "depth_unit": 1, "fov_range_relative": [0.5, 1.0], "image_augmentation": ["jittering", "jpeg_loss", "blurring", "shot_noise", "dof"] }, { "name": "Synscapes", "path": "path/to//Synscapes/", "label_type": "A", "weight": 2.0, "depth_unit": 1, "fov_range_relative": [0.5, 1.0], "image_augmentation": ["jittering", "jpeg_loss", "blurring", "shot_noise", "dof"] }, { "name": "UnrealStereo4K", "path": "path/to//UnrealStereo4K/", "label_type": "A", "weight": 1.7, "depth_unit": 1, "fov_range_relative": [0.33, 1.0], "image_augmentation": ["jittering", "jpeg_loss", "blurring", "shot_noise", "dof"] }, { "name": "EDEN", "path": "path/to//EDEN/", "label_type": "A", "weight": 1.2, "fov_range_relative": [0.5, 1.0], "image_augmentation": ["jittering", "jpeg_loss", "blurring", "shot_noise", "dof"] } ] }, "model_version": "v2", "model": { "encoder": { "backbone": "dinov2_vitl14", "intermediate_layers": [5, 11, 17, 23], "dim_out": 1024 }, "neck": { "dim_in": [1026, 2, 2, 2, 2], "dim_out": null, "dim_res_blocks": [1024, 256, 128, 64, 32], "num_res_blocks": [0, 2, 2, 2, 0], "res_block_in_norm": "none", "res_block_hidden_norm": "none", "resamplers": ["conv_transpose", "conv_transpose", "conv_transpose", "bilinear"] }, "points_head": { "dim_in": [1024, 256, 128, 64, 32], "dim_out": [null, null, null, null, 3], "dim_res_blocks": [1024, 256, 128, 64, 32], "num_res_blocks": [0, 1, 1, 1, 0], "res_block_in_norm": "none", "res_block_hidden_norm": "none", "resamplers": ["conv_transpose", "conv_transpose", "conv_transpose", "bilinear"] }, "normal_head": { "dim_in": [1024, 256, 128, 64, 32], "dim_out": [null, null, null, null, 3], "dim_res_blocks": [1024, 256, 128, 64, 32], "num_res_blocks": [0, 1, 1, 1, 0], "res_block_in_norm": "none", "res_block_hidden_norm": "none", "resamplers": ["conv_transpose", "conv_transpose", "conv_transpose", "bilinear"] }, "mask_head": { "dim_in": [1024, 256, 128, 64, 32], "dim_out": [null, null, null, null, 1], "dim_res_blocks": [1024, 256, 128, 64, 32], "num_res_blocks": [0, 1, 1, 1, 0], "res_block_in_norm": "none", "res_block_hidden_norm": "none", "resamplers": ["conv_transpose", "conv_transpose", "conv_transpose", "bilinear"] }, "scale_head": { "dims": [1024, 1024, 1024, 1] }, "remap_output": "exp", "num_tokens_range": [1200, 3600] }, "optimizer": { "type": "AdamW", "params": [ {"params": {"include": ["*"], "exclude": ["*.backbone.*"]}, "lr": 1e-4}, {"params": {"include": ["*.backbone.*"]}, "lr": 1e-5} ] }, "lr_scheduler": { "type": "SequentialLR", "params": { "schedulers": [ {"type": "LambdaLR", "params": {"lr_lambda": ["1.0", "max(0.0, min(1.0, (epoch - 1000) / 1000))"]}}, {"type": "StepLR", "params": {"step_size": 25000, "gamma": 0.5}} ], "milestones": [2000] } }, "low_resolution_training_steps": 50000, "loss": { "invalid": {}, "A": { "global": {"function": "affine_invariant_global_loss", "weight": 1.0, "params": {"align_resolution": 48}}, "patch_4": {"function": "affine_invariant_local_loss", "weight": 1.0, "params": {"level": 4, "align_resolution": 24, "num_patches": 16}}, "patch_16": {"function": "affine_invariant_local_loss", "weight": 1.0, "params": {"level": 16, "align_resolution": 12, "num_patches": 256}}, "patch_64": {"function": "affine_invariant_local_loss", "weight": 1.0, "params": {"level": 64, "align_resolution": 6, "num_patches": 4096}}, "normal": {"function": "edge_loss", "weight": 1.0}, "normal_map": {"function": "normal_map_loss", "weight": 0.1}, "metric_scale": {"function": "metric_scale_loss", "weight": 0.1}, "mask": {"function": "mask_bce_loss", "weight": 0.1} }, "B": { "global": {"function": "affine_invariant_global_loss", "weight": 1.0, "params": {"align_resolution": 48}}, "patch_4": {"function": "affine_invariant_local_loss", "weight": 1.0, "params": {"level": 4, "align_resolution": 24, "num_patches": 16}}, "patch_16": {"function": "affine_invariant_local_loss", "weight": 1.0, "params": {"level": 16, "align_resolution": 12, "num_patches": 256}}, "metric_scale": {"function": "metric_scale_loss", "weight": 0.1}, "normal": {"function": "edge_loss", "weight": 1.0}, "normal_map": {"function": "normal_map_loss", "weight": 0.1}, "mask": {"function": "mask_bce_loss", "weight": 0.1} }, "C": { "global": {"function": "affine_invariant_global_loss", "weight": 1.0, "params": {"align_resolution": 48}}, "patch_4": {"function": "affine_invariant_local_loss", "weight": 1.0, "params": {"level": 4, "align_resolution": 24, "num_patches": 16}}, "metric_scale": {"function": "metric_scale_loss", "weight": 0.1}, "mask": {"function": "mask_bce_loss", "weight": 0.1} } } }