| # dataset-related | |
| raw_data_dir: data/raw/videos | |
| processed_data_dir: data/processed/videos | |
| binary_data_dir: data/binary/videos | |
| video_id: '' | |
| task_cls: '' | |
| # project-related | |
| work_dir: '' | |
| load_ckpt: '' | |
| tb_log_interval: 100 | |
| num_ckpt_keep: 1 | |
| val_check_interval: 10000 | |
| valid_infer_interval: 10000 | |
| num_sanity_val_steps: 0 | |
| num_valid_plots: 5 | |
| eval_max_batches: 100 # num_test_plots | |
| print_nan_grads: false | |
| resume_from_checkpoint: 0 # specify the step, 0 for latest | |
| amp: false | |
| valid_monitor_key: val_loss | |
| valid_monitor_mode: min | |
| save_best: true | |
| debug: false | |
| save_codes: | |
| - tasks | |
| - modules | |
| - egs | |
| # testing related | |
| gen_dir_name: '' | |
| save_gt: true | |
| # training-scheme-related | |
| max_updates: 40_0000 | |
| seed: 9999 | |
| lr: 0.0005 | |
| scheduler: exponential # exponential|rsqrt|warmup|none|step_lr | |
| warmup_updates: 0 | |
| optimizer_adam_beta1: 0.9 | |
| optimizer_adam_beta2: 0.999 | |
| weight_decay: 0 | |
| clip_grad_norm: 0 # disable grad clipping | |
| clip_grad_value: 0 # disable grad clipping | |
| rays_sampler_type: uniform | |
| in_rect_percent: 0.95 | |
| accumulate_grad_batches: 1 | |
| # model-related | |
| use_window_cond: true | |
| with_att: true # only available when use win_cond, use a attention Net in AD-NeRF | |
| cond_type: '' | |
| cond_dim: 64 | |
| hidden_size: 256 | |
| # NeRF-related | |
| near: 0.3 | |
| far: 0.9 | |
| n_rays: 1600 # default 2048, 1600 for RTX2080Ti | |
| n_samples_per_ray: 64 | |
| n_samples_per_ray_fine: 128 | |
| embedding_args: | |
| multi_res_pos: 10 # log2+1 of max freq for positional encoding (3D location) | |
| multi_res_views: 4 # log2+1 of max freq for positional encoding (2D direction) | |
| infer_cond_name: '' | |
| infer_out_video_name: '' | |
| infer_scale_factor: 1.0 | |
| infer_smo_std: 0. | |
| infer_audio_source_name: '' | |
| infer_c2w_name: '' | |
| # postprocessing params | |
| infer_lm3d_clamp_std: 1.5 | |
| infer_lm3d_lle_percent: 0.25 # percent of lle fused feature to compose the processed lm3d | |
| infer_lm3d_smooth_sigma: 0. # sigma of gaussian kernel to smooth the predicted lm3d | |
| infer_pose_smooth_sigma: 2. | |
| load_imgs_to_memory: false # load uint8 training img to memory, which reduce io costs, at the expense of more memory occupation |