Shulou(Shulou.com)06/01 Report--

How to use the multi-person attitude recognition framework AlphaPose. In order to solve this problem, this article introduces the corresponding analysis and solution in detail, hoping to help more partners who want to solve this problem to find a more simple and feasible method.

A brief introduction to Multi-person attitude recognition

There are two mainstream schemes for human posture estimation: Two-step framework and Part-based framework. The first scheme is to detect each human body detection box in the environment, and then independently detect the posture of each human body region (top-down method). The second scheme is to first detect all the limb nodes in the environment, and then splice to get multiple skeletons (bottom-up method). In the first scheme, the quality of the target area frame detection since the height of attitude detection accuracy. In the second scheme, if the two people are very close, it is easy to be ambiguous, and because it depends on the relationship between the two components, it loses access to the overall information.

AlphaPose

AlphaPose uses a top-down approach and proposes a RMPE (Regional Multi-person attitude Detection) framework. The framework mainly includes symmetric spatial transformer network (SSTN), Parametric Pose Non- Maximum-Suppression (NMS) and Pose-Guided Proposals Generator (PGPG). And three techniques, symmetric spatial transformer network (SSTN), deep proposals generator (DPG) and parametric pose nonmaximum suppression (p-NMS), are used to solve the problem of multi-person attitude estimation in field scenes.

By adding SSTN to the SPPE structure, high-quality human body regions can be extracted from imprecise region boxes. Parallel SPPE branches (SSTN) to optimize its own network. Parametric pose NMS is used to solve the problem of redundancy detection. In this structure, a self-created attitude distance measurement scheme is used to compare the similarity between gestures. The attitude distance parameters are optimized by data-driven method. Finally, we use PGPG to strengthen the training data, by learning the description information of different postures in the output results, to imitate the generation process of the human body region box, and further generate a larger training set.

Follow the repo installation instructions to complete the installation of the code base, and then run the following command line:

Python scripts/demo_inference.py-cfg configs/coco/resnet/256x192_res50_lr1e-3_1x.yaml-checkpoint pretrained_models/fast_res50_256x192.pth-indir examples/demo/

The running results are as follows:

This is the answer to the question on how to use the multi-person attitude recognition framework AlphaPose. I hope the above content can be of some help to you. If you still have a lot of doubts to be solved, you can follow the industry information channel for more related knowledge.

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.