How to used in the customer dataset?

[jo-dean]

Hello,i want use the model in my customer dataset. But the result is not so good just used the provided model in this hub.
Any advice to fit my own dataset?
Thank you.

[hetao828]

Do you solve this problem? When I train my novel class 'zangxueji' the new test dataset shows that 'zangxueji' instances is 0? Why? and it;s train will be dead soon.

[hetao828]

[hetao828]

I think this problem maybe caused by 0 instances of 'zangxueji', because sometimes it shows CUDA OUT OF MEMORY.

[Assass1no]

Hello,i want use the model in my customer dataset. But the result is not so good just used the provided model in this hub. Any advice to fit my own dataset? Thank you.

请问在自己的数据集上用DeFRCN需要修改代码的哪些部分？是按照dectron2的数据集导入来修改吗？

[aeinsara]

Hello. I have this problem. Do you solve it?

[aeinsara]

Do you solve this problem? When I train my novel class 'zangxueji' the new test dataset shows that 'zangxueji' instances is 0? Why? and it;s train will be dead soon.

I found the reason of this problem.It because of this line in "defrcn/dataloader/build.py":
"histogram += np.histogram(classes, bins=hist_bins)[0]".
bins=hist_bins must be adapted to yours dataset. For example my classes is [13,14,15,16], so bins should be [13,14,15,16,17] for display correctly.
This problem only exists in the display and in fact, the code works correctly. So do not worry.

[hui-yiming]

你好，我想请问一下为什么我换做自己的数据集之后，loss等一起切训练正常，但是最后的检测结果却是0 呢？

[orzgugu]

你好，我想请问一下为什么我换做自己的数据集之后，loss等一起切训练正常，但是最后的检测结果却是0 呢？

Hello, I would like to ask how you changed to your own data set for training and testing, I am very confused about this, I am sorry to bother you, I am an AI rookie.