Can I train on the basis of maskrcnn's pre-trained model? If the categories of the pre-trained model are different from those of my own dataset, will the pre-trained model still be effective?
我可以按照Maskrcnn预先训练的模式进行训练吗?如果预先训练的模型的类别与我自己的数据集的类别不同,那么预先训练的模型是否仍然有效?
I tried with a new class training dataset. But it seems to converge very slowly.
我尝试了一个新的班级训练数据集。但它似乎收敛得非常慢。
更多回答
I've given a very broad answer to your question. You'll have to provide more details if you want more specific answers. What is your task? What is your dataset? On which dataset was the model pretrained?
我已经对你的问题给出了一个非常宽泛的答案。如果你想要更具体的答案,你必须提供更多的细节。你的任务是什么?您的数据集是什么?模型是在哪个数据集上预先训练的?
Please clarify your specific problem or provide additional details to highlight exactly what you need. As it's currently written, it's hard to tell exactly what you're asking.
请澄清您的具体问题或提供更多详细信息,以突出您的确切需求。按照目前的写法,很难准确地说出你在问什么。
优秀答案推荐
Generally, fine-tuning pretrained models on your own dataset is a well established practice. It is common to replace the model's head with your own that will match the number of classes in your dataset and will have randomly initialized weights at first.
通常,在您自己的数据集上微调预先训练的模型是一种公认的做法。通常将模型的头部替换为您自己的头部,该头部将与数据集中的类数匹配,并且一开始将具有随机初始化权重。
更多回答
我是一名优秀的程序员,十分优秀!