3D Human Keypoints Estimation From Point Clouds in the Wild Without Human Labels

TL;DR

This paper introduces GC-KPL, an unsupervised method for estimating 3D human keypoints from point clouds without requiring labeled data, leveraging geometry consistency to achieve competitive results.

Contribution

The paper presents a novel unsupervised learning approach for 3D human keypoint detection from point clouds, reducing reliance on expensive annotations.

Findings

Achieves reasonable keypoint estimation performance without human labels.

Enhances downstream few-shot learning with minimal labeled data.

Outperforms state-of-the-art methods when trained on entire dataset.

Abstract

Training a 3D human keypoint detector from point clouds in a supervised manner requires large volumes of high quality labels. While it is relatively easy to capture large amounts of human point clouds, annotating 3D keypoints is expensive, subjective, error prone and especially difficult for long-tail cases (pedestrians with rare poses, scooterists, etc.). In this work, we propose GC-KPL - Geometry Consistency inspired Key Point Leaning, an approach for learning 3D human joint locations from point clouds without human labels. We achieve this by our novel unsupervised loss formulations that account for the structure and movement of the human body. We show that by training on a large training set from Waymo Open Dataset without any human annotated keypoints, we are able to achieve reasonable performance as compared to the fully supervised approach. Further, the backbone benefits from the unsupervised training and is useful in downstream fewshot learning of keypoints, where fine-tuning on only 10 percent of the labeled training data gives comparable performance to fine-tuning on the entire set. We demonstrated that GC-KPL outperforms by a large margin over SoTA when trained on entire dataset and efficiently leverages large volumes of unlabeled data.