Jiazhong Cen
Jiazhong Cen
你好，这里看上去只是文件名错误，请查看./logs/llff\fern\这个文件夹找到对应的tar文件。在我印象里这个文件不是这样命名的，你是否修改了代码中的某些内容呢？
能否提供下你之前使用的具体脚本，指令等，目前这些信息我们没办法debug
This [one](https://github.com/Jumpat/SegmentAnythingin3D/tree/sa3d-gs).
Hi! Did you change any code? The '2356.25 GiB' to be allocated seems a little weird. This seems to be caused by an unexpected broadcast operation.
> Which command should be used if running MipNeRF-360 dataset？llff/kitchen.py or nerf_unbounded/kitchen.py > I don't really understand You may need to use the nerf_unbounded config as the MIP360 dataset involves...
Have you run the run.py successfully to get the pertained NeRF model? Maybe you can check whether the reload_ckpt_path has the corresponding NeRF model (like fine_last.tar).
I guess this is caused by the missing 'c' in the config file 'seg_kitchen.py'. I mean the expname = 'dvgo_kitchen_unbounded' should be 'dcvgo_kitchen_unbounded'.
Hi! You can find the code about NeRF in lib/dvgo.py (dcvgo, seg_dvgo, ...) > hello，I would like to ask where the whole model framework is and how to understand it....
There is no batch size in SA3D. You can reduce the resolution of TensoRF (mask grids resolution, TensoRF grids resolution, rendering resolution, density grids resolution, ...) for saving memory.
> hello, I would like to ask if the parameters in the code are already optimal? Do you still need hyperparameter optimization? For some scenes and targets it is. However...