search

zubair-irshad / shapo

Pytorch code for ECCV'22 paper. ShAPO: Implicit Representations for Multi-Object Shape, Appearance and Pose Optimization
Other
181 stars 12 forks source link

Problems with running code #7

Closed dongho-Han closed 1 year ago

dongho-Han commented 1 year ago

Hello. Thanks for the nice work! I tried to run your 'ShAPO' code, but I got stuck. Can you share the solution how to solve the problems written below?

  1. As mentioned in #6 , there is no auto_encoder_model folder and sdf_rgb_pretained/rgb_latent folder in supported download link. Can you update the link?

  2. When I try to run the distributed script, python prepare_data/distributed_generate_data.py --data_dir /home/ubuntu/shapoplusplus/data/nocs_data --type camera_train I got an error message like this FileNotFoundError: [Errno 2] No such file or directory: '/data/sdf_rgb_pretrained/rgb_net_weights/reconstructor.pt' Because there is no rgb_net_weights/reconstructor.pt in $ShAPO_Repo/data. How to solve this?

  3. After download and organize a small Real test subset, pretrained checkpoints,... directory shape is like this test_data ├── ckpts │   └── shapo_real.ckpt ├── Real │   ├── test │   │   ├── scene_1 │   │   │   ├── 0006_color.png │   │   │   └── 0006_depth.png │   │   ├── scene_2 │   │   │   ├── 0049_color.png │   │   │   └── 0049_depth.png │   │   └── scene_3 │   │   ├── 0003_color.png │   │   └── 0003_depth.png │   └── test_list_subset.txt ├── sdf_rgb_pretrained │   ├── all_train_ids.json │   ├── LatentCodes │   │   ├── 1000.pth │   │   ├── 100.pth │   │   ├── 2000.pth │   │   ├── 500.pth │   │   ├── all_train_ids.json │   │   └── latest.pth │   ├── Logs.pth │   ├── ModelParameters │   │   ├── 1000.pth │   │   ├── 100.pth │   │   ├── 2000.pth │   │   ├── 500.pth │   │   └── latest.pth │   ├── OptimizerParameters │   │   ├── 1000.pth │   │   ├── 100.pth │   │   ├── 2000.pth │   │   ├── 500.pth │   │   └── latest.pth │   ├── rgb_net_weights │   │   ├── feats_colors.pt │   │   └── reconstructor.pt │   └── specs.json

It means that there is no sdf_rgb_pretrained/Reconstructions, sdf_rgb_pretrained/rgb_latent in $ShAPO_Repo/test_data.

Is there any missing points that I'm doing?

  1. Also, is there a way to inference with pre-trained weights without training?

Thanks!

zubair-irshad commented 1 year ago

Hi @dongho-Han,

Thanks for your interest in our work. Please see the following answer https://github.com/zubair-irshad/shapo/issues/6#issuecomment-1382195978 for your question 1.2 and 3.

For question 4. Yes, absolutely. Please see this notebook where we demo inference with pretrained weights without re-training. You can also check number 3. here as we provide a dedicated script inference_real here.

Hope it helps!