yuxumin
commented
2 years ago Hi,
This is because that we pass a wrong key to shapenet_dict ('./data/shapenet_synset_dict.json').
If you want to see the reconstruction metric for each class in ModelNet40, you need first:
- Modify https://github.com/lulutang0608/Point-BERT/blob/master/datasets/ModelNetDataset.py#L149 as the same as https://github.com/lulutang0608/Point-BERT/blob/master/datasets/ShapeNet55Dataset.py#L67. Return the right taxonomy id of each model, you can define a self-design
taxonomy idaccording to the label of each sample. - Create a
taxonomy : namedict. As ./data/shapenet_synset_dict.json
Or if you do not care about the detailed metric. Comment out https://github.com/lulutang0608/Point-BERT/blob/master/tools/runner.py#L271-L295
sw778
JohanYe
Hello, I try to use ModelNet40 for pre-training of dVAE. I changed the dataset_name in runner.py to ModelNet40 and modified the code as shown in the picture. In addition, I also configured the dvae.yaml file about ModelNet40. But when I start training I get "KeyError: 'ModelNet'". I don't know why this problem occurs, is it related to the json file? , or do you have any good suggestions? Thanks.
`(pointbert) ws666@ws666-OMEN-by-HP-Laptop-15-dc1xxx:~/Point-BERT-master$ bash scripts/train.sh 0 --config cfgs/ModelNet_models/dvae.yaml --exp_name modelnetDvae
Traceback (most recent call last): File "main.py", line 72, in