Yuzhong Zhao

Sorry for the delay, we have uploaded a simple demo code. By running `python demo/dynrefer_inference.py` , object informations can be predicted. You can use different models and images by modifying...

Sorry for the delay, we have uploaded a simple demo code. By running `python demo/dynrefer_inference.py` , object informations can be predicted. You can use different models and images by modifying...

Most of the packages can be downloaded following `https://github.com/callsys/DynRefer/blob/main/docs/install.md`. Other missing packages can be downloaded by `pip`.

I think cuda 11.7 is fine.

Can you provide the error message. The enviroment of the code is similar to LAVIS. I think installing torch and torchvision that matches the version of coda 11.7 can be...

Sorry for the delay response, `bert-base-uncased` can be downloaded by tools like `huggingface_hub`.

你好，表1中使用的测试集对应（configs/base/datasets/vg/vg_reg.yaml）这个文件，可以在[Glamm](https://drive.google.com/file/d/1zF3UGHU1rvgTujinqJ-hZtrCBVsfsuel/view?usp=sharing)这下载，也可以使用我们提供的链接下载我们处理过的文件。

你好！没有尝试过与这一篇比较过精度

可能存在一定不公平比较的问题。regiongpt似乎没有开源代码，无法确定使用的测试集。grit使用了vg数据集的所有测试样本。我们follow的是Glamm/Groma/Osprey这几篇文章，使用5000个bbox来做测试。

是的 2025-04-14 19:12:57"liweiyangxjtu" ***@***.***>写道： 你好，表1中使用的测试集对应（configs/base/datasets/vg/vg_reg.yaml）这个文件，可以在Glamm这下载，也可以使用我们提供的链接下载我们处理过的文件。 请问这个是用vg1.2训练的模型测试吗 — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you commented.Message ID: ***@***.***> liweiyangv left a...