BOSHU LEI

I'd like to suggest your guys to check the original code of [LSS](https://github.com/nv-tlabs/lift-splat-shoot), We re-implement it as a FoV to BeV encoder and get satisfying results.

@hskAlena How do you choose the initial views ?

OK. We use every 5 images in the first 20 images in the training split. ("./train/r_0", "./train/r_5", "./train/r_10", "./train/r_15"). I think that's the reason we produce different results.

We train for 200k iterations. Test skip is 8. We pick one every 8 images in test set. The total number is 25.

@hskAlena Hi, have you reproduced the results in the paper ?

Hi, we want to reproduce the results in your Table 2. From the caption, Setting I is initally trained with 4 observations. I wonder how you choose these 4 observations....

I think you miss some files in the dataset -- `Frames.m4v`, you can check whether this file is downloaded correctly

you should generate them. please set the `redo` in `configs/preprocess/sfm_spp_spg_sample.yaml` to True. It applies to other split (train / test / val).

Hi, I encounter a similar problem. I can't find file `NodeStatus.hpp`. I find the same module `libuavcan` in PX4 repo Firmware and I can build it. So I believe the...