Question about 3D Localization Mechanism

[Xiaolong-RRL]

Dear author:

Thanks for your interesting work.

I have two questions about 3D Localization Mechanism:

1. Are position embeddings already embedded in the features of 3dllm_final_scene_data_v2
2. Where in the code is the embedding of location tokens implemented? I can not find it.

By the way, I wander when will you release detail data & code about grounding? I am really looking forward to it~

Best！
Xiaolong

[cazhang]

Also interested in how the location tokens are used for 3d grounding task?

• For now, the code shows how it is initialised only. However, it is not clear how the token is mapped to 3d space.
• For example, the output may be like: the sofa is at <location 12>, <location 1>, <location 37>. But what does this mean? Each location corresponding to XYZ position, or a scalar number ?

Hope the author could explain more on this!

Thanks. @evelinehong

[1768637871]

Also interested in how the location tokens are used for 3d grounding task?还对位置令牌如何用于 3d 接地任务感兴趣？

• For now, the code shows how it is initialised only. However, it is not clear how the token is mapped to 3d space.目前，代码仅显示它的初始化方式。但是，目前尚不清楚令牌如何映射到 3D 空间。
• For example, the output may be like: the sofa is at <location 12>, <location 1>, <location 37>. But what does this mean? Each location corresponding to XYZ position, or a scalar number ?例如，输出可能如下：沙发位于<位置 12>、<位置 1>、<位置 37>。但这意味着什么呢？每个位置对应于 XYZ 位置，还是一个标量数？

Hope the author could explain more on this!希望作者能对此进行更多解释！

Thanks.  谢谢。@evelinehong

Hello, how can you get the grounding outputs? Could you tell me the format of your prompts? I couldn't find it in the paper or in the projects. Thanks a lot!