vobecant / POP3D

Source code for NeurIPS paper "POP-3D: Open-Vocabulary 3D Occupancy Prediction from Images"
https://vobecant.github.io/POP3D/
92 stars 7 forks source link

Regarding zeroshot_weights.pth #12

Closed jjeong19 closed 2 months ago

jjeong19 commented 2 months ago

Hi,

Thank you for your work on language driven occupancy prediction! I had a question regarding the text embeddings used. There only seem to be approximately 32 classes for nuScenes, however the embedding is of dimensions [102, 512] instead of [32, 512]. Would it be possible for you to provide text prompts used to generate the text embeddings?

Sincerely, Jason

vobecant commented 2 months ago

Dear Jason,

thank you very much for your question. Please see Table 2 in our paper https://arxiv.org/abs/2401.09413 for the categories' descriptions.

Thanks, Antonin.