用于图像分类的神经网络 (OpenCV/Emgu)
我是神经网络的新手，并且已经查看了 emgu 和论坛上的（稀疏）代码示例。但是无法弄清楚如何使用 OpenCV 中的神经网络实现对图像进行分类。
例如
- 类别：车辆（子类别：汽车、摩托车、卡车等），
建筑物：房屋、摩天大楼、小屋等），
人群：男性、女性等） 动物：狗、猫、老虎等。
有人可以提供一些代码示例或指针吗？非常感谢您的帮助。
I am new to neural networks and have looked over the (sparse) code samples on emgu and on the forum. However could not figure out how to classify a images using the neural network implementation in OpenCV.
e.g.
- Categories: vehicles (subcategories: cars, motorcycles, trucks etc.),
buildings: houses, skyscrapers, huts etc.),
people: men, women etc.)
animals: dogs, cats, tigers etc.
Could someone provide some code samples or pointers? Your help is much appreciated.
如果你对这篇内容有疑问，欢迎到本站社区发帖提问 参与讨论，获取更多帮助，或者扫码二维码加入 Web 技术交流群。
绑定邮箱获取回复消息
由于您还没有绑定你的真实邮箱，如果其他用户或者作者回复了您的评论，将不能在第一时间通知您！
发布评论
评论（3）
您的任务超出了图像分类的当前技术水平。练习一些非常非常简单的东西，例如数字识别。
另外，请检查 ASIRRA 项目中的猫/狗分类问题。
Your task is beyond the current state of the art in image classification. Practice with something much, much simpler, like digit recognition.
Also, check the ASIRRA project for the cat/dog classification problem.
如果您选择尝试数字识别，这里是我在两门课程中介绍的数据集用于神经网络的第一个作业。
If you choose to get your feet wet with digit recognition, here's the dataset that both courses I took used for the first assignment in Neural Networks.
完全同意 Don Reba。提供更多细节，您很可能没有足够的数据来解决您正在尝试的问题。事实上，斯图尔特·格曼（Stuart Geman）在过去几年的演讲中认为，认为分类可以通过这种方式“解决”视觉是不现实的。它被称为 “Google 和风险投资维度”。
通过大量的数据，数字识别和这种“规模”的问题更容易解决。
Concurring completely with Don Reba. Giving more specifics, you most likely have not nearly enough data to solve the problems you're trying. In fact, Stuart Geman gave a talk within the last few years that thinking classification can "solve" vision in this way is not realistic. It was called "Google and the VC Dimension".
Digit recognitions and problems of that "size" are more solvable with a decent amount of data.