ocnflag / point2seq

Point2Seq

This is a reproduced repo of Point2Seq for 3D object detection.

The code is mainly based on OpenPCDet.

Introduction

We provide code and training configurations of PointSeq on the ONCE and Waymo Open dataset under tools/cfgs.

Requirements

The codes are tested in the following environment:

• Ubuntu 18.04
• Python 3.6
• PyTorch 1.5
• CUDA 10.1
• OpenPCDet v0.3.0
• spconv v1.2.1

Installation

a. Clone this repository.

b. Install the dependent libraries as follows:

• Install the dependent python libraries:

pip install -r requirements.txt 

• Install the SparseConv library, we use the implementation from [spconv].
  • If you use PyTorch 1.1, then make sure you install the spconv v1.0 with (commit 8da6f96) instead of the latest one.
  • If you use PyTorch 1.3+, then you need to install the spconv v1.2. As mentioned by the author of spconv, you need to use their docker if you use PyTorch 1.4+.

c. Compile CUDA operators by running the following command:

python setup.py develop

Training

All the models are trained with Tesla V100 GPUs (32G). If you use different number of GPUs for training, it's necessary to change the respective training epochs to attain a decent performance.

point2seq's Issues

no huawei_dataset.yaml

hello, could you please provide cfgs/dataset_configs/huawei_dataset.yaml? many thanks.

center_ops_cuda

Hi, ocNflag. Thanks for your works. I was training the code and there was an ImportError: cannot import name 'center_ops_cuda' from 'pcdet.ops.center_ops' (unknown location). I wonder if there was something I missed.

outdated code

Code doesn't support latest pytorch versions anything above 1.10

where is the cfgs/dataset_configs/waymo_dataset.yaml?

Hello, Thank you for your great works.
I have a problem when I want to use tools/cfg/waymo_p2s.yaml.
it refer the cfgs/dataset_configs/waymo_dataset.yaml but there isn't this file.

I would be very appreciate if you could share this file.
Thank you :)

Question about the DCNV2Fusion Module

Hi,

Thanks for your awesome work and open-sourced code!
I am trying to understand the principle behind the DCNV2Fusion Module, which is not mentioned in your paper.
Can you please give me suggestions like other paper to help me understand the fusion among center_map, foreground_map and corner_map?