neutral model

[ghumpire]

Hi,

I ran the demo but got the assertion indicating that the data/smpl/SMPL_NEUTRAL.pkl does not exist. Then I noticed the fetch data section indicates that the 'neutral model' (http://smplify.is.tue.mpg.de/) is needed for training, evaluation, and running the demo code. That website contains multiple files, Could you please confirm if that model is the one located on lsp_results.targ.gz (obtained from http://smplify.is.tue.mpg.de/) > results > lsp > all_results_gender-neutral.pkl? If so, I would have to rename all_results_gender-neutral.pkl as SMPL_NEUTRAL.pkl. After renaming the pkl, the code returned the error below, which makes me think I got the wrong pkl:

File "/multiperson/mmdetection/mmdet/models/utils/smpl/smpl.py", line 51, in __init__
    super(SMPL, self).__init__(*args, create_global_orient=False, create_body_pose=False, create_betas=False, create_transl=False, **kwargs)
  File "/python3.6/site-packages/smplx/body_models.py", line 205, in __init__
    self.faces = data_struct.f
AttributeError: 'Struct' object has no attribute 'f'

Cheers, Gabriel

[manlio99]

Hey i had a related mistake, i didn't solve it yet. i think that I'm forgetting something.... can u help me?

console:

/home/massiris/miniconda3/envs/multiperson/bin/python3.7 /home/massiris/Desktop/python/3dpose/multiperson/mmdetection/tools/demo.py unexpected key in source state_dict: fc.weight, fc.bias

missing keys in source state_dict: layer2.0.bn2.num_batches_tracked, layer2.2.bn3.num_batches_tracked, layer2.2.bn1.num_batches_tracked, layer2.0.downsample.1.num_batches_tracked, layer3.0.bn3.num_batches_tracked, layer1.1.bn1.num_batches_tracked, layer4.1.bn3.num_batches_tracked, layer2.1.bn2.num_batches_tracked, layer2.2.bn2.num_batches_tracked, layer3.0.bn1.num_batches_tracked, layer4.0.bn2.num_batches_tracked, layer4.2.bn1.num_batches_tracked, layer3.5.bn3.num_batches_tracked, layer3.4.bn1.num_batches_tracked, layer4.2.bn2.num_batches_tracked, layer3.1.bn2.num_batches_tracked, layer1.1.bn2.num_batches_tracked, layer3.1.bn1.num_batches_tracked, layer4.0.downsample.1.num_batches_tracked, layer1.0.bn3.num_batches_tracked, layer3.1.bn3.num_batches_tracked, layer2.3.bn1.num_batches_tracked, layer3.4.bn2.num_batches_tracked, layer3.2.bn2.num_batches_tracked, layer3.4.bn3.num_batches_tracked, layer3.2.bn3.num_batches_tracked, layer3.3.bn2.num_batches_tracked, layer4.1.bn1.num_batches_tracked, layer4.0.bn1.num_batches_tracked, layer2.0.bn3.num_batches_tracked, layer1.2.bn2.num_batches_tracked, bn1.num_batches_tracked, layer3.3.bn3.num_batches_tracked, layer3.5.bn1.num_batches_tracked, layer1.1.bn3.num_batches_tracked, layer2.0.bn1.num_batches_tracked, layer3.3.bn1.num_batches_tracked, layer1.0.downsample.1.num_batches_tracked, layer4.2.bn3.num_batches_tracked, layer1.2.bn3.num_batches_tracked, layer2.1.bn3.num_batches_tracked, layer4.0.bn3.num_batches_tracked, layer2.3.bn3.num_batches_tracked, layer4.1.bn2.num_batches_tracked, layer2.3.bn2.num_batches_tracked, layer3.0.bn2.num_batches_tracked, layer3.2.bn1.num_batches_tracked, layer1.0.bn1.num_batches_tracked, layer3.5.bn2.num_batches_tracked, layer1.0.bn2.num_batches_tracked, layer1.2.bn1.num_batches_tracked, layer3.0.downsample.1.num_batches_tracked, layer2.1.bn1.num_batches_tracked

2020-06-22 16:09:39,883 - INFO - load checkpoint from /home/massiris/Desktop/python/3dpose/multiperson/mmdetection/data/checkpoint.pt 2020-06-22 16:09:40,274 - WARNING - missing keys in source state_dict: smpl_head.smpl.lbs_weights, smpl_head.loss.smpl.v_template, smpl_head.smpl.parents, smpl_head.loss.smpl.faces_tensor, smpl_head.loss.smpl.posedirs, smpl_head.smpl.posedirs, smpl_head.loss.smpl.vertex_joint_selector.extra_joints_idxs, smpl_head.loss.smpl.shapedirs, smpl_head.loss.smpl.parents, smpl_head.loss.smpl.J_regressor_extra, smpl_head.smpl.J_regressor, smpl_head.smpl.v_template, smpl_head.smpl.faces_tensor, smpl_head.smpl.vertex_joint_selector.extra_joints_idxs, smpl_head.loss.smpl.lbs_weights, smpl_head.smpl.shapedirs, smpl_head.smpl.J_regressor_extra, smpl_head.loss.smpl.J_regressor

Traceback (most recent call last): File "/home/massiris/Desktop/python/3dpose/multiperson/mmdetection/tools/demo.py", line 199, in main() File "/home/massiris/Desktop/python/3dpose/multiperson/mmdetection/tools/demo.py", line 158, in main runner.resume(cfg.resume_from) File "/home/massiris/miniconda3/envs/multiperson/lib/python3.7/site-packages/mmcv/runner/runner.py", line 313, in resume self.optimizer.load_state_dict(checkpoint['optimizer']) File "/home/massiris/miniconda3/envs/multiperson/lib/python3.7/site-packages/torch/optim/optimizer.py", line 115, in load_state_dict raise ValueError("loaded state dict contains a parameter group " ValueError: loaded state dict contains a parameter group that doesn't match the size of optimizer's group

Process finished with exit code 1

[sleep3r]

@ghumpire, @manlio99

Docker image I used - nvidia/cudagl:10.0-devel-ubuntu16.04

Server preparations

apt-get update
apt-get install sudo git wget unzip -y
sudo apt-get install build-essential libglib2.0-0 libsm6 libxext6 libxrender-dev -y
sudo apt-get install libglfw3-dev libgles2-mesa-dev -y

In case you have no anaconda installed

wget https://repo.anaconda.com/archive/Anaconda3-2020.02-Linux-x86_64.sh
bash Anaconda3-2020.02-Linux-x86_64.sh

Some checks

gcc --version
nvcc -V
conda

git clone https://github.com/JiangWenPL/multiperson.git

cd multiperson/
conda env create -f environment.yml
source activate multiperson

cd neural_renderer/
python3 setup.py install

cd ../mmcv
python3 setup.py install

cd ../mmdetection
./compile.sh

python setup.py develop

cd ../sdf
python3 setup.py install

Data fetching

cd ../mmdetection/
mkdir data
mkdir results
cd data

pip install gdown

gdown https://drive.google.com/uc?id=1y5aKzW9WL42wTfQnv-JJ0YSIgsdb_mJn
tar xf multiperson_data.tar -C ~/multiperson/mmdetection/data

wget https://people.eecs.berkeley.edu/~kanazawa/cachedir/hmr/neutral_smpl_mean_params.h5 --no-check-certificate

Downloading the SMPL model from my Google Drive

gdown https://drive.google.com/uc?id=<someId>
unzip smpl.zip

Downloading neutral model .pkl from smplify_code_v2.zip

cd smpl
gdown https://drive.google.com/uc?id=<anotherId>

According to #1

rm -rf /root/anaconda3/envs/multiperson/lib/python3.7/site-packages/mmcv*
cd ../mmcv
python3 setup.py install

Now can run it

cd ~/multiperson/mmdetection/
python3 tools/demo.py --config=configs/smpl/tune.py --image_folder=demo_images/ --output_folder=results/ --ckpt data/checkpoint.pt

[Anirudh257]

@eladwass I didn't perform the upgrade from Python 2 to Python 3 but I followed the steps outlined by @sleep3r and my code ran.

[eladwass]

Hi

I successfully downloaded : "Downloading neutral model .pkl from smplify_code_v2.zip"

But what exactly do u download here: "Downloading the SMPL model from my Google Drive" - I took the PKL from the models folder inside the smplify_code_v2.zip

I performed the steps @sleep3r did but tackled the following: ImportError: cannot import name 'ModelOutput' from 'smplx.body_models'

To overcome this I replaced the above line with: from smplx.body_models import SMPLXOutput in "/multiperson/mmdetection/mmdet/models/utils/smpl/smpl.py"

[seyeeet]

@eladwass Thank you! It helps me a lot! Can you tell me what it does when you replace it with SMPLOutput? what is the difference and how did you figure it out :1st_place_medal:

[Dingpx]

@sleep3r I can't find the data you outlined.

1. SMPL model 2. neutral model .pkl Could you tell me how to find it?

[JiangWenPL]

Should be resolved in the latest updates.