Break through the dimensional barriers and let the dolls around you move on the screen!

Because of love, people will make the two-dimensional image into a doll figure, but the cold model without life breath cannot meet people's emotional needs for interaction. How to make doll figures with emotional sustenance more expressive and appealing?

Recently, the HMS Core 3D modeling service has launched the automatic bone binding capability, which allows the bipedal humanoid objects that have been modeled to move according to custom actions, and even interact with users. It is no longer just a cold and rigid doll figure.

So, what can 3D models that are animated by automatic bone binding capabilities be used for?

For example, using audio and video editing capabilities , you can build 3D models of children's favorite dolls, add dance movements, dubbing and children's song stories, and edit them into relatively complete children's teaching videos to help children receive knowledge faster and better;

For another example, combined with the action production capabilities of 3D Engine , smooth 3D animation effects can be achieved through functions such as real-time skeletal animation, facial expression driving, full-body IK, and multi-animation state machine fusion, coupled with 3D Engine's high-definition rendering and visual effects. , intelligent wayfinding navigation and other functions, you can make a mature animation game;

For another example, with AR Engine 's AR capabilities such as motion tracking, body and face tracking, and environment tracking, 3D models can be placed in real scenes under the lens and interact with users. This scene can be applied to the field of AR games to realize the function of customizing 3D models and model actions in games, making AR interactive games more interesting.

In addition, we can break through the imagination and add two "legs" to the bottles and objects in life during the modeling process, so that the automatic bone binding ability is no longer limited to specific model categories. Everything can be activated, truly realizing the interesting scene of "recovery of all things".

Implementation principle

The automatic bone binding service is aimed at developers who need 3D games and 3D animation production. By inputting a static 3D model of a bipedal humanoid object, limbs can be bound based on AI algorithms, and the model's bones and skeletons can be automatically generated. Skin weights, complete automatic bone binding, and then dynamically drive the model to complete custom actions such as walking, jumping, and dancing by changing the orientation and position of the bones, so as to give the 3D model the ability to move.

Ability advantage

1. One-key binding skinning can be achieved without manual intervention.

Bone binding is divided into manual binding and automatic binding. Most of the current high-precision automatic bone binding methods require that the model be in a standard pose, and seven or eight key points must be manually placed before binding. Huawei's automatic skeleton binding capability is based on AI intelligent algorithms. It only needs to keep the model in a roughly correct position, and does not even require the model to be absolutely upright and facing forward. The model can be accurately bound without manually placing the key point.

2. Based on massive data, improve the accuracy and generalization of AI binding algorithm.

So, how does Huawei's automatic bone binding capability make 3D model data to ensure the accuracy of the AI binding algorithm? Different from the time-consuming and labor-intensive traditional 3D model data creation method, the 3D model data creation of Huawei's automatic skeleton binding capability only needs to use ordinary mobile phones to capture and collect object images, and automatically generate 3D character data through Huawei's self-developed algorithm. When training algorithms on hundreds of thousands of 3D model-bound data generated on a large scale, only a small amount of real data needs to be fine-tuned to obtain an AI model with high accuracy and generalization.

Enter model constraints

Considering the diversification of 3D world models, in order to ensure the accuracy of the automatic bone binding algorithm, we make the following suggestions for the input model of the binding algorithm:

Source : Photo scan modeling using bipedal humanoid figures (excluding hand-held objects) and plush toys;

Appearance : From a visual point of view, there is no separation of the grid, no adhesion between the limbs and the body, no large appendages, try to stand with both legs and arms outstretched;

Attitude : The overall posture is that the z-axis faces forward and the y-axis faces upwards, and each joint can have a disturbance of no more than 15°, and no symmetry is required;

Mesh : Mesh is a triangular surface or a quadrilateral surface, the number of vertices should not exceed 80k, and the model as a whole has no large-scale missing patches;

Others : The proportions of the body and limbs conform to the style of most toys, the limbs should not be too thin or too short (the ratio of the width of arms and legs should not be less than 8% of the longest side of the model), and the overall posture of the model is straight forward.

Huawei's HMS Core 3D modeling service 's automatic bone binding capability, through intelligent technology traction, greatly reduces the threshold for 3D model animation production, animation production is no longer a difficult technical ability that requires complex equipment, even if non-professionals use ordinary The mobile phone can also easily complete animation production, and realize more creative work production in daily life.

Learn more details>>

Visit the official website of Huawei Developer Alliance
Get development guidance documents
Huawei Mobile Services Open Source Warehouse Address: GitHub , Gitee

Follow us to know the latest technical information of HMS Core for the first time~