{ "model": { "name": "PatchFusion", "version_name": "patchfusion", "n_bins": 64, "bin_embedding_dim": 128, "bin_centers_type": "softplus", "n_attractors":[16, 8, 4, 1], "attractor_alpha": 1000, "attractor_gamma": 2, "attractor_kind" : "mean", "attractor_type" : "inv", "midas_model_type" : "DPT_BEiT_L_384", "min_temp": 0.0212, "max_temp": 50.0, "output_distribution": "logbinomial", "memory_efficient": true, "inverse_midas": false, "img_size": [384, 512], "do_resize": false, // do resize in dataloader to speed up "raw_depth_shape": [2160, 3840], "sr_ratio": 1, "sampled_training": false, "transform_sample_gt_size": [2160, 3840], "sample_feat_level": 4, "use_hr": false, "baseline": true, "condition": true, "freeze": true, "g2l": true, "use_fusion_network": true, "use_area_prior": true }, "train": { "use_rgb": true, "use_blur": false, "train_midas": true, "use_pretrained_midas": true, "trainer": "zoedepth_custom", "epochs": 12, "bs": 16, "optim_kwargs": {"lr": 0.0001, "wd": 0.01}, "sched_kwargs": {"div_factor": 10, "final_div_factor": 10000, "pct_start": 0.5, "three_phase":false, "cycle_momentum": true}, "same_lr": false, "w_si": 1, "w_domain": 0.2, "w_reg": 0, "w_grad": 0, "avoid_boundary": false, "random_crop": true, "input_width": 640, "input_height": 480, "midas_lr_factor": 1, "encoder_lr_factor":10, "pos_enc_lr_factor":10, "freeze_midas_bn": true, "sec_stage": true, "multi_consistency": true }, "infer":{ "train_midas": false, "use_pretrained_midas": false, "pretrained_resource" : "url::https://github.com/isl-org/ZoeDepth/releases/download/v1.0/ZoeD_M12_N.pt", "force_keep_ar": true }, "eval":{ "train_midas": false, "use_pretrained_midas": false, "pretrained_resource" : "url::https://github.com/isl-org/ZoeDepth/releases/download/v1.0/ZoeD_M12_N.pt" } }