Bottom-up Human Pose Estimation

hello, how many gpus did you used? I found that your batchsize maybe is different from the original HigherHRNet, can you share the batchsize you used for training? Thanks a lot!

When I run python tools/dist_inference.py \ --img_dir path/to/your/directory/of/images \ --save_dir path/where/results/are/saved \ --cfg experiments/coco/higher_hrnet/w32_512_adam_lr1e-3.yaml \ TEST.MODEL_FILE models/pose_coco/pose_higher_hrnet_w32_512.pth \ TEST.SCALE_FACTOR '[0.5, 1.0, 2.0]' It occurs that RuntimeError: Attempting to deserialize object on a CUDA device but torch.cuda.is_available() is False. If you are running on a CPU-only machine, please use torch.load with map_location=torch.device('cpu') to map your storages to the CPU. Besides, there are another problem Process Process-1: Traceback (most recent call last): File "/home/ubuntu/anaconda3/envs/mykaolin/lib/python3.7/multiprocessing/process.py", line 297, in _bootstrap self.run() File "/home/ubuntu/anaconda3/envs/mykaolin/lib/python3.7/multiprocessing/process.py", line 99, in run self._target(*self._args, **self._kwargs) File "tools/dist_inference.py", line 123, in worker model.load_state_dict(torch.load(cfg.TEST.MODEL_FILE), strict=True) File "/home/ubuntu/anaconda3/envs/mykaolin/lib/python3.7/site-packages/torch/nn/modules/module.py", line 1052, in load_state_dict self.__class__.__name__, "\n\t".join(error_msgs))) RuntimeError: Error(s) in loading state_dict for PoseHigherResolutionNet: Missing key(s) in state_dict: "conv_scale_layers.0.0.weight", "conv_scale_layers.0.0.bias", "conv_scale_layers.0.1.weight", "conv_scale_layers.0.1.bias", "conv_scale_layers.0.1.running_mean", "conv_scale_layers.0.1.running_var", "conv_scale_layers.0.3.weight", "conv_scale_layers.0.3.bias", "conv_scale_layers.1.0.weight", "conv_scale_layers.1.0.bias", "conv_scale_layers.1.1.weight", "conv_scale_layers.1.1.bias", "conv_scale_layers.1.1.running_mean", "conv_scale_layers.1.1.running_var", "conv_scale_layers.1.3.weight", "conv_scale_layers.1.3.bias". I am confused why GPU doesn't work.

I get this error when i try to use dist_inference , seems like there is a file missing or did I failed the installation?

Traceback (most recent call last):
  File "/home/kubler/data/HPE/SWAHR-HumanPose/tools/dist_inference.py", line 42, in <module>
    from utils.vis import draw_skeleton
ImportError: cannot import name 'draw_skeleton' from 'utils.vis' (/home/kubler/data/HPE/SWAHR-HumanPose/tools/../lib/utils/vis.py)```


Thanks for your help!

Greg

As we can see, the scale network branch directly regresses alpha = 1/s-1. When using Taylor expansion as shown in Eq. 8, in fact, due to sigma_0 has changed to ssigma_0, the coverage of gaussian should change from 3sigma to 3s* sigma, where is not hold in Eq. 8.

Thanks for releasing the code. By comparing SWAHR-HumanPose and HigherHRNet, I notice that some nn.ReLU layers are added in the fuse_layers. I wonder what if we do not modify the original backbone network. Will it affect the final performance?

My Environment is Ubuntu 18.04LTS, pytorch-1.10, RTX2080Ti, cuda 11.0.

Traceback (most recent call last): File "tools/dist_train.py", line 320, in main() File "tools/dist_train.py", line 115, in main args=(ngpus_per_node, args, final_output_dir, tb_log_dir) File "/home/lab/.conda/envs/pytorch-1.10/lib/python3.7/site-packages/torch/multiprocessing/spawn.py", line 230, in spawn return start_processes(fn, args, nprocs, join, daemon, start_method='spawn') File "/home/lab/.conda/envs/pytorch-1.10/lib/python3.7/site-packages/torch/multiprocessing/spawn.py", line 188, in start_processes while not context.join(): File "/home/lab/.conda/envs/pytorch-1.10/lib/python3.7/site-packages/torch/multiprocessing/spawn.py", line 150, in join raise ProcessRaisedException(msg, error_index, failed_process.pid) torch.multiprocessing.spawn.ProcessRaisedException:

-- Process 0 terminated with the following error: Traceback (most recent call last): File "/home/lab/.conda/envs/pytorch-1.10/lib/python3.7/site-packages/torch/multiprocessing/spawn.py", line 59, in _wrap fn(i, *args) File "/media/hdd/xxy/swahr/tools/dist_train.py", line 161, in main_worker rank=args.rank File "/home/lab/.conda/envs/pytorch-1.10/lib/python3.7/site-packages/torch/distributed/distributed_c10d.py", line 576, in init_process_group store, rank, world_size = next(rendezvous_iterator) File "/home/lab/.conda/envs/pytorch-1.10/lib/python3.7/site-packages/torch/distributed/rendezvous.py", line 183, in _tcp_rendezvous_handler store = _create_c10d_store(result.hostname, result.port, rank, world_size, timeout) File "/home/lab/.conda/envs/pytorch-1.10/lib/python3.7/site-packages/torch/distributed/rendezvous.py", line 158, in _create_c10d_store hostname, port, world_size, start_daemon, timeout, multi_tenant=True RuntimeError: Address already in use

In section 3.2, you introduced the scale-adaptive heatmaps, but in formula 4, you write s.t. ||i-x||<3σ ||j-y||<3σ, dose this σ refer to σ0 or σ0s? At the first, i thought this σ refer to σ0s, but in formula 6, you write when Hσ0 == 0, Hσ0s=Hσ0, which means the region whose value is not zero in Hσ0 and Hσ0s is the same, it's inconsistent with what you said "which means that the region convered by this Gaussian kernel will also become larger", so what are the facts?

Dear author: How many GPUs is used during training? How many days does it take to train for 300 epochs? I have 8 GPUs, but the training seems to take 3 weeks... Thank you!

RuntimeError: Attempting to deserialize object on a CUDA device but torch.cuda.is_available() is False. If you are running on a CPU-only machine, please use torch.load with map_location=torch.device('cpu') to map your storages to the CPU.

my torch.cuda.is_available() return true,why it can not work .thanks

When using the code to train the model using the crowd pose dataset without any changes in the code, why is the AP 62.7, which is not consistent with the result (AP=71.6) presented in this paper?