xheon/jointembedding

[ICCV'19] Joint Embedding of 3D Scan and CAD Objects

jointembedding's Issues

Issue about ScanNet file

Thank you for your great world.
I wonder if you could give us some instructions about preprocess scannet file. For example, the input format of scannet is '.mask', what's the content of it.
Great thanks for your help

Pre-trained weights?

Hi,
Can you provide the pre-trained weights?

Scan2Cad training data not available

Thanks for your great work.
I have some questions about the training data preparation.
I followed the Scan2CAD repository and download its annotation dataset.
However, it seems that there's no training, validation split within it. Besides, there's also no field called 'scan2cad_objects' in the dataset's json file. This is also not included in the json file of the Scan-CAD Object Similarity dataset. Which file should be filled in the field of "--scan2cad_file"?
I wonder if you can give us some more instructions on the training data preparation.
Thank you very much.

Data preprocessing

Hello, thank you for your contribution. Could you please provide the code for data preprocessing

Can not reproduce the results in the paper when training from scratch

Hi Manuel,
thanks for your work. I am having trouble to reproduce the retrieval accuracies (or ranking accuracies) quoted in the paper and in issue #4. I am training the network from scratch with the provided code (without any modifications) on the provided preprocessed data. The results are so far off that it seems like there must be a bug somewhere.

I get the following total retrieval accuracies after the respective number of train steps.
Number steps: 2k , 25 k, 50 k , 75 k , 100 k
Orig settings repo Retrieval accuracy: 0.03, 0.09, 0.10, 0.08, 0.03
Mod settings Retrieval accuracy: 0.01 , 0.03, 0.03, 0.05 , 0.05

For the mod settings in line with the paper I set the triplet margin to 0.2 (as opposed to 0.01) and change the learning rate scheduler to update every 20 k steps as opposed to 40 k steps. As far as I can tell all other settings in the repo match the information provided in the paper.

The result quoted in the paper is 0.43 and in issue #4 it is 0.42.
I know this is fairly old code but would be amazing if either you or @TimFelixBeyer could look into this.
Thanks a lot!
Florian

where is the code

Test .py file for pre-trained weight

hello,thanks for your excellent work, i got some trouble to apply the pre-trained weight you given, but the file got bug(pre-trained weight have dismatched dimension).
Can you provide the test.py which is suitable for the pre-trained weight? Thanks a lot.