In 21st century, deep learning/machine learning is a buzz word and almost every area of computing is trying to add in this new component in. It is interesting to see learning methods are used in animation aspect as well. Most of the papers are published in recent years and there are different results for character control specifically. The learning bicycle stunts paper is an interesting one. The team has presented an approach to simulate and control a human character riding a bike. The rider not only learns to steer and to balance in normal riding situations, but also learns to perform a wide variety of stunts. The rider is optimized through offline learning. From the result, you could see the rider is able to balance while doing some difficult stunts, however, I feel like it is a little bit unrealistic or too prefect in the real world. The learning method normally are introduced to get a more efficient simulation with less iterations of calculation, while applying the learning methods, I think sometimes for a lot of different application of learning method, it is important to evaluate the usability and availability instead of just following the trend.
Link to bicycle stunt paper: https://dl.acm.org/citation.cfm?id=2601121&picked=formats
0 Comments
Leave a Reply. |
The site will mainly share about what I learned and found interesting in Technical animation class.Categories |