dichen-cd / hoim-pytorch Goto Github PK

I am very fortunate to read your paper.

I would like to know how to use OIM loss to train the network？

Thank you for your answer.

I would like to ask what is the difference between this code and your other NAE?

Hi! Thank you so much for the nicely written paper and clean implementation!
I was wondering if there is a way to load a few frames without having any pre-existing gallery with its corresponding targets that contain the box coordinates and labels. Would the network be able to cut the people out of said frames and eventually re-identify them a few frames later?

非常有幸拜读您的论文，感觉很受启发。在训练模型的过程中，我遇到了一些小问题，还希望您能慷慨解答。

I'm trying to train model with 3 RTX 2080Ti using:
我尝试用三张2080Ti进行训练，使用的命令和参数是：

CUDA_VISIBLE_DEVICES=0,1,2 python -m torch.distributed.launch --nproc_per_node=3 scripts/train_hoim.py --debug --lr_warm_up -p ./logs/self-train-batch6_c/ --batch_size 2 --nw 5 --w_RCNN_loss_bbox 10.0 --epochs 22 --lr 0.003 --distributed.

I encountered some issue with the SyncBatchNorm which said expect at least 3D input (got 2D input). According to my investigation, SyncBatchNorm is not compatible with BatchNorm1d having a 2D input in PyTorch 1.2. So I substitute the BatchNorm1d in faster_rcnn_hoim.py with a workaround:
由于PyTorch 1.2版本的SyncBatchNorm与BatchNorm1d不是很兼容，不支持2D的输入，所以我自己写了一个类来替换faster_rcnn_hoim.py中的BatchNorm1d：

class MyBatchNorm1d(nn.Module):

    def __init__(self, *args):
        super(MyBatchNorm1d, self).__init__()
        self.bn2d = nn.BatchNorm2d(*args)

    def forward(self, x):
        x = x[..., None, None]
        x = self.bn2d(x)
        return x[..., 0, 0]

However, I end up with inferior performance even compared with a model trained with batch=2 on a single GPU.
但是最终使用分布式训练得到的结果比单卡、batch=2训练出来的结果还要差。

I noticed you used a single GPU in the experiment of your paper, I'm wondering if you encountered the similar issue or I did not configured the code correctly.
我注意到在您的论文中使用了单张显卡进行训练，我想知道您是否遇到了相似的问题，亦或是我没有我的训练方式有误？

Thanks!
谢谢！

There may be something wrong with this file. I can't install most of the libraries in the file.
I tried to install in multiple versions of conda, and they all had the same error.
So I would like to ask how to use this file and whether it is the wrong document?

Have you ever tried to add FPN structure into your network?