search

Arthur151 / ROMP

Monocular, One-stage, Regression of Multiple 3D People and their 3D positions & trajectories in camera & global coordinates. ROMP[ICCV21], BEV[CVPR22], TRACE[CVPR2023]
https://www.yusun.work/
Apache License 2.0
1.33k stars 229 forks source link

Run video inference encoutered error #292

Closed zeng-yifei closed 2 years ago

zeng-yifei commented 2 years ago

Hello! ROMP is really a great work, and your document makes everything so easy to use. But when I followed the step in google colab to install ROMP in my local machine, and run

python -m romp.predict.image --inputs=demo/images --output_dir=demo/image_results --renderer=pytorch3d

then I encountered bug report as:

No configs_yml is set, set it to the default --configs_yml=configs/image.yml yaml_timestamp /home/gpu/content/ROMP/active_configs/active_context_2022-07-03_15_21_17.yaml Loading the configurations from configs/image.yml Traceback (most recent call last): File "/usr/local/anaconda3/lib/python3.8/runpy.py", line 194, in _run_module_as_main return _run_code(code, main_globals, None, File "/usr/local/anaconda3/lib/python3.8/runpy.py", line 87, in _run_code exec(code, run_globals) File "/home/gpu/content/ROMP/romp/predict/image.py", line 79, in <module> main() File "/home/gpu/content/ROMP/romp/predict/image.py", line 71, in main processor = Image_processor(args_set=args_set) File "/home/gpu/content/ROMP/romp/predict/image.py", line 14, in __init__ super(Image_processor, self).__init__(**kwargs) File "/home/gpu/content/ROMP/romp/predict/base_predictor.py", line 12, in __init__ super(Predictor, self).__init__(**kwargs) TypeError: __init__() got an unexpected keyword argument args_set'

then I print the args_set, it looks like this:

{'args_set': Namespace(FOV=60, GPUS=0, HMloss_type='MSE', PCK_thresh=150, Rot_typ e='6D', Synthetic_occlusion_ratio=0, add_offsetmap=True, adjust_lr_epoch=[], adju st_lr_factor=0.1, baby_threshold=0.8, backbone='hrnet', batch_size=64, beta_dim=1 0, bv_with_fv_condition=True, calc_PVE_error=False, calc_mesh_loss=True, calc_pck =False, calc_smpl_mesh=True, cam_dim=3, cam_dist_thresh=0.1, center_def_kp=True, centermap_conf_thresh=0.25, centermap_size=64, character='smpl', collect_subdirs= False, collision_aware_centermap=False, collision_factor=0.2, color_jittering_rat io=0.2, configs_yml='configs/image.yml', dataset='h36m,mpii,coco,aich,up,ochuman, lsp,movi', dataset_rootdir='/home/gpu/content/dataset/', deconv_num=0, depth_loss _type='Piecewise', distillation_learning=False, distributed_training=False, epoch =120, eval=False, eval_2dpose=False, eval_datasets='pw3d', f=None, fast_eval_iter =-1, fine_tune=True, fix_backbone_training_scratch=False, focal_length=443.4, fps _save=24, fv_conditioned_way='attention', gpu='0', head_block_num=2, homogenize_p ose_space=False, hrnet_pretrain='/home/gpu/content/ROMP/trained_models/pretrain_h rnet.pkl', image_loading_mode='image', input_size=512, inputs='demo/images', inte ractive_vis=False, kernel_sizes=[5], learn_2dpose=False, learn_AE=False, learn_kp 2doffset=False, learn_relative=False, learn_relative_age=False, learn_relative_de pth=False, learn_relative_shape=False, local_rank=0, log_path='/home/gpu/content/ log/', loss_thresh=1000, lr=0.0003, make_tracking=False, master_batch_size=-1, ma tch_preds_to_gts_for_supervision=False, matching_mode='all', matching_pckh_thresh =0.6, max_person=64, max_supervise_num=-1, merge_smpl_camera_head=False, mesh_clo th='ghostwhite', model_path='trained_models/ROMP_HRNet32_V1.pkl', model_precision ='fp32', model_return_loss=False, model_version=1, multi_person=True, new_trainin g=False, num_depth_level=8, nvxia_model_path='/home/gpu/content/ROMP/model_data/c haracters/nvxia', nw=4, optimizer_type='Adam', output_dir='demo/image_results', p erspective_proj=False, print_freq=50, renderer='pytorch3d', resnet_pretrain='/hom e/gpu/content/ROMP/trained_models/pretrain_resnet.pkl', rot_dim=6, rotate_prob=0. 2, sampling_aggregation_way='floor', save_centermap=False, save_dict_results=True , save_mesh=True, save_visualization_on_img=True, scale_anchor=True, show_largest _person_only=False, show_mesh_stand_on_image=False, shuffle_crop_mode=False, shuf fle_crop_ratio_2d=0.1, shuffle_crop_ratio_3d=0.9, smil_model_path='/home/gpu/cont ent/ROMP/model_data/smpl_models/smil_packed_info.pth', smpl_joint_num=22, smpl_me sh_root_align=False, smpl_model_path='/home/gpu/content/ROMP/model_data/smpl_mode ls/smpl_packed_info.pth', smpl_prior_path='/home/gpu/content/ROMP/model_data/para meters/gmm_08.pkl', smpl_uvmap='/home/gpu/content/ROMP/model_data/parameters/smpl _vt_ft.npz', smpla_model_path='/home/gpu/content/ROMP/model_data/smpl_models/SMPL A_NEUTRAL.pth', soi_camera='far', supervise_cam_params=False, supervise_global_ro t=False, surrounding_camera=False, tab='hrnet_cm64_process_images', teacher_model _path='/export/home/suny/CenterMesh/trained_models/3dpw_88_57.8.pkl', temporal_op timization=False, test_interval=2000, top_n_error_vis=6, track_memory_usage=False , use_coordmaps=True, use_eft=True, val_batch_size=4, voc_dir='/home/gpu/content/ dataset/VOCdevkit/VOC2012/', wardrobe='/home/gpu/content/ROMP/model_data/wardrobe ', weight_decay=1e-06)}

I have totally no idea about why this occurs. Hope for your answer :)

Arthur151 commented 2 years ago

Sorry about the bug. The inference usage of original version of ROMP has been out of data. The google colab would be updated later. For inference, please use simple-romp. https://github.com/Arthur151/ROMP/tree/master/simple_romp The rest code is just used for training.