SEUvictor

about requirements.txt

2

Why is requirements.txt so much and so messy, and what does '@ file...' mean?  Can you release a simple and effective version？

编译失败

3

DCN和iou3d都无法编译成功，把pytorch换成1.7，DCN也换成适合1.7的版本后，可以编译，但是iou3d还是无法编译。 但是iou3d不编译也能正常运行，请问iou3d在哪起作用呢？ 另外可以更新一下版本吗，pytorch1.0有点低了...

hey,guy why can't you provide a pretrained weight?

about self.norm

1

Hi， Thanks for your excellent work. Why does the regularization method in the model use nn.InstanceNorm2d instead of BN?

Hi,Thanks for your wonderful job, I have a question, Is it enough to use centernet for 3D target detection training only with image information? Or does it need camera parameters?

请教一些问题，感谢

2

您好，在查找多目标跟踪方案的时候看到了您的工作，感觉很棒很感兴趣，想学习一下然后试着用到自己的论文里。 还不太懂目标跟踪任务，有几个问题想请教一下，希望能够被解答，谢谢~ 1.您是用yolov7官方提供的预训练模型，在VisDrone2019-MOT数据集上训练的吗？ 2.这是只再训练了检测器yolov7呢，还是同时训练了跟踪器呢？比如bytetrack，这个跟踪器本身有需要训练学习的参数吗？ 3.您提供的模型权重是只训练了10 epoch的吗？ 4.weights文件夹下的DHN.pth和ckpt.t7两个权重文件是什么的呢？跟踪器的吗？

The strack_pool list is predicted before matching, Why does the unconfirmed list not need to be predicted by Kalman filter before matching? strack_pool列表在匹配前都有先进行预测，为什么unconfirmed列表在匹配前不需要经过卡尔曼滤波进行预测？

How can I get my own 'database. txt', 'train. txt' and 'test. txt' when training my own dataset? What is the relationship between them? For example, are 'train. txt' and...

关于多卡训练卡死的问题

1

对代码进行了修改以训练自己数据，发现多卡训练总是不能进行下去，单卡是没问题。调试发现是卡在了下面这一步。 `msg = model.load_state_dict(checkpoint, strict=False)`  请问是为什么呢？ 训练命令为： `CUDA_VISIBLE_DEVICES=0,1 python -m torch.distributed.run --nproc_per_node 2 --master_port 12345 main.py`