JDHR

Introduction (介绍)

JDHR是一个基于Jittor国产深度学习框架的动态人体渲染算法库。该算法库全面集成了动态人体渲染的关键技术，包括点云采样、4D特征网格表示以及实时渲染等多个关键模块。

JDHR (Jittor-based Dynamic Human Rendering) is a dynamic human rendering algorithm library based on Jittor. This algorithm library fully integrates key technologies, including point cloud sampling, 4D feature grid representation, and real-time rendering.

Plan (开源计划)

• Release training code
• Release High frame rate(HFR) rendering code
• Release initializing point clouds code
• Build a WebSocket-based viewer

Installation（安装）

Install the basic environment under the JDHR repo:

# Editable install, with dependencies from requirements.txt
pip install -v -e . 

Install Rasterizer for realtime rendering:

cd easyvolcap/diff_point_rasterizater
mkdir build && cd build
cmake .. -DCMAKE_CXX_COMPILER=g++ -DCMAKE_CUDA_ARCHITECTURES=86(根据显卡版本选用70.75.86)
make -j4

Datasets（数据集）

DNA-Rendering Datasets

Please refer to HumanRF to download DNA-Rendering datasets. Note that you should cite the corresponding papers if you use these datasets.

🕒 more datasets

training（训练）

This script trains a single-frame version on the first frame of the 0013_09 sequence of the DNA-Rendering dataset. You can quickly verify whether your dataset preparation and installation process are correct.

evc-train -c configs/exps/4k4d/4k4d_0013_09_r4.yaml,configs/specs/static.yaml,configs/specs/tiny.yaml exp_name=4k4d_0013_09_r4_static

The actual training of the full model is more straight forward:

evc-train -c configs/exps/4k4d/4k4d_0013_09_r4.yaml

During the validation phase, rendering frame rate should be greater than 30 FPS

🕒HFR Rendering（高帧率渲染）

Team （团队）

动态人体渲染算法库（JDHR）是由清华大学和北京交通大学团队共同维护的开源代码库。欢迎大家使用JDHR开展研究工作。

JDHR is an open-source code repository jointly maintained by teams from Tsinghua University and Beijing Jiaotong University

Feel free to request support for new models and contribute to JDHR.

Acknowledgement (鸣谢)

1. Jittor
2. JNeRF
3. EasyVolcap
4. 3DGS