A Simple Baseline for Efficient
Hand Mesh Reconstruction
(CVPR 2024)

Zhishan Zhou*, Shihao Zhou*, Zhi Lv, Minqiang Zou, Yao Tang and Jiajun Liang
* Equally Contribution. ✝ Corresponding Author.
JIIOV Technology

[Paper]     [Github]     [BibTeX]

Click to play the demo video. An alternate Bilibili source is HERE.

JIIOV Technology is a leading biometric recognition solutions provider. We are now working on Extended Reality (XR) hand interaction algorithm and modules.
We transplanted simpleHand design into realworld device to make a practically test. It shows advantages especially in finger interaction senario, while retains realtime efficiency. Additional videos showcasing a comparison between our method and both Quest and AVP will be made available soon.


We propose simpleHand, a simple yet effective baseline that not only surpasses state-of-the-art (SOTA) methods but also demonstrates computational efficiency. SimpleHand can be easily transplant to mainstream backbones and datasets.
SimpleHand consists a backbone and a mesh decoder, while mesh decoder is abstracted into a token generator and a mesh regressor.
From functional perspective, Token generator samples representative tokens using predicted 2d keypoints. Mesh regressor cascadely lifts the sampled tokens into meshes. We proposed a concept of "CORE STRUCTURE", which is the minimum structure that meets these functions.

Results and Comparisons

SimpleHand found the component with the greatest performance gain and implemented them using the simplest structure. Therefore, it has good performance in both efficiency and accuracy.
Comparison to existing work
SimpleHand capitalizes on the strengths of existing methodologies, thereby outperforming them in numerous challenging scenarios.This is particularly evident in intricate finger interactions, like pinching or twisting, as well as in complex and unconventional gestures.

About us

Founded in February 2020, JIIOV Technology is a consumer electronics provider specializing in fingerprint solutions, smart components and innovative sensor technology. Within one year of our establishment, our products had been successfully applied in leading smartphone companies devices. We owe this achievement to our employees commitment to our core value of "pioneering innovation"

Our mission

Creating a brand new interactive experience with AI and sensors, enable the world to accelerate into a fully intelligent era.

Our vision

Creating an era of safe, user-friendly, convenient and smart interaction.

Through the ongoing efforts of all JIIOV employees, we confidently provide you with better, smarter products and experiences.
— Chen Keqing


  title={A Simple Baseline for Efficient Hand Mesh Reconstruction},
  author={Zhishan Zhou and Shihao. zhou and Zhi Lv and Minqiang Zou and Yao Tang and Jiajun Liang},

Other Works

We won the 1st place award at The 7th Workshop on Observing and Understanding Hands in conjunction with ICCV 2023, AssemblyHands Track. This challenge aims at reconstructing egocentric 3D hand joints in a video sequence.
Our technical report is available HERE.



We would like to express our heartfelt gratitude to 于陌尘, 吴桐, 金日强, 黄怡菲, 邹佳辰, 丁岩, 张雅婷, 李杰, 牟理慧 and 杨梅 for their invaluable support in various aspects of this project. Their contributions have been instrumental in shaping the algorithm's ideas, developing the demo application, deploying on mobile platforms, constructing imaging camera modules, and conducting performance tests. Each of them has brought unique expertise and insights that have significantly enhanced the project's quality and success.

Project page template is borrowed from DreamBooth.