search

nickgkan / butd_detr

Code for the ECCV22 paper "Bottom Up Top Down Detection Transformers for Language Grounding in Images and Point Clouds"
Other
74 stars 11 forks source link

Performance about the Group-Free 3D object detector trained to detect 485 object categories in ScanNet #25

Closed ZCMax closed 1 year ago

ZCMax commented 1 year ago

Thanks for your paper and code, I've got a question about the Group-Free 3D object detector trained to detect 485 object categories in ScanNet, can you offer the pre-trained checkpoints or some related per-class performance?

ayushjain1144 commented 1 year ago

Hi, here are the pre-trained weights: https://drive.google.com/file/d/1JwMTOaMWfK0JgOBBHU_2oBGXp9ORo9Q3/view?usp=sharing

We also provide the box outputs in the ReadMe, maybe you could use it to calculate its mAP (from what I vaguel remember, it was around 58% computed on all 485 classes. but this was long time back, so i am not sure)