Open b10232031 opened 4 years ago
Hi, actually, the shape for prediction is (B, N, 13, 13, A, (5+1)). In RegionLoss the number of classes is actually 1 instead of N, as shown in the config file athttps://github.com/bingykang/Fewshot_Detection/blob/20c0580356e273cd69ac16ae327226649419b44b/cfg/darknet_dynamic.cfg#L259.
Thanks for your correction, I'll try it later.
Hi, actually, the shape for prediction is (B, N, 13, 13, A, (5+1)). In RegionLoss the number of classes is actually 1 instead of N, as shown in the config file at
.
So this means that model can only predict only 1 class? Because the model I retrain can only detect 1 class.
The result is strange, I wonder that the model can only predict only one class ?
Hi, actually, the shape for prediction is (B, N, 13, 13, A, (5+1)). In RegionLoss the number of classes is actually 1 instead of N, as shown in the config file at https://github.com/bingykang/Fewshot_Detection/blob/20c0580356e273cd69ac16ae327226649419b44b/cfg/darknet_dynamic.cfg#L259
.
So this means that model can only predict only 1 class? Because the model I retrain can only detect 1 class.
Hi,could you please tell me which cfg file do you use while running detect.py,I just got so many bounding box in one picture and any of them are wrong bounding boxes!!
The result is strange, I wonder that the model can only predict only one class ?
您好 武汉的前辈 我跑该源码时我遇到了如下问题 这个问题前面也有人提到过 但并未得到解答 您可以解决吗?有偿 “ ”
The result is strange, I wonder that the model can only predict only one class ?
您好 武汉的前辈 我跑该源码时我遇到了如下问题 这个问题前面也有人提到过 但并未得到解答 您可以解决吗?有偿 “ ”
我当时也遇到了相同的问题,原因是我更变了gpu的数量,一定要在配置文件中修改gpu的数量,比如有两块gpu,就是ngpu=0,1
Sorry for troubling you, but I don't know how to compute loss for meta-model and feature-extractor.
In my idea, we'll have predict vector with shape (B, N, 13, 13, A, (5+N)) after feature reweighting, where B is batch, N is classes and A is anchors. If so, should I split my ground truth to N vectors according to different classes and compute loss for each channel of N in predict vector? And the second question is the loss for meta-model and feature-extractor is the same one?
I fondly anticipate your reply, thanks.