Jinbin Fu

I have the same comfusion, why the code use size_average and batch_average at the same time, it seems divide the batch_size twice?

> Hi, thanks for your amazing work. In the paper, you mentioned using lambda_1 and lambda_2, which balancing weights to the two tasks. However, I found self.s_det and self.s_id are...

> @AllenPu 你好，你使用py3.6复现成功了吗？我使用py3.6的时候在**train/provider.py**中`import cPickle as pickle`的时候，显示py3不支持cPickle，于是我换成pickle但是在加载数据的时候出现了错误`Traceback (most recent call last): File "train/train.py", line 68, in rotate_to_center=True, random_flip=True, random_shift=True, one_hot=True) File "/home/cy/F-PointNet/frustum-pointnets-master/train/provider.py", line 142, in __init__ self.box2d_list = pickle.load(fp) UnicodeDecodeError: 'ascii'...

> Also bothered me for quite a long time. Solved by compiling with **gcc-11 and g++-12**. Hope this helps you ^_^ > > * CUDA Version: 11.8 > * OS:...

> Hi, could you show me your custom dataset config (whole exported config in exp folder is better) and OOM output? These are config and error log Config: weight =...

I random split the train and val, so there won't be much difference. You can see before OOM, the code try to allocate 5.38 GB, in my custom dataset, each...

> I am sorry about that issue. I never encounter a similar problem. I notice that the validation batch size per GPU is identical to the train batch size per...