Long-Tailed-Recognition.pytorch icon indicating copy to clipboard operation
Long-Tailed-Recognition.pytorch copied to clipboard

[NeurIPS 2020] This project provides a strong single-stage baseline for Long-Tailed Classification, Detection, and Instance Segmentation (LVIS). It is also a PyTorch implementation of the NeurIPS 202...

Results 26 Long-Tailed-Recognition.pytorch issues
Sort by recently updated
recently updated
newest added

Hi, when i run "python main.py --cfg ./config/CIFAR10_LT/causal_norm.yaml", i got this error: RuntimeError: Given input size: (2048x1x1). Calculated output size: (2048x-5x-5). Output size is too small but when i run...

Hey @KaihuaTang, I am an NLPer. Thanks for your interesting work. I've observed that you got the same cofigs regarding the parameters in class Causal_Norm_Classifier, namely num_head, tau, alpha and...

作者你好,最近才看到这篇关于长尾数据分类的论文。有一些地方不是很明白。 1)论文指出验证或测试时才需要消除头部偏移的影响,那么消除头部偏移影响后,模型在训练集上的表现如何? 2)该偏移可以看作零输入的特征输出,那么能否将网络中所有卷积和BatchNormalization的bias均设置为否? 3) alpha的含义是什么,为什么可以大于1 ![image](https://user-images.githubusercontent.com/21170245/156131305-e60950b7-d7b2-4cd0-9e20-298fda314685.png) 谢谢

您好,首先非常感谢您solid的工作。 但在我看具体代码时,有一些问题想要请教: 在 lvis1.0\mmdet\models\roi_heads\bbox_heads\convfc_bbox_head.py 文件中的 update_embed 函数里: ``` def update_embed(self, targets, gt_label): if self.training: # remove background with torch.no_grad(): fg_target = targets[gt_label > 0].clone().detach().mean(0, keepdim=True) self.causal_embed = self.MU * self.causal_embed...

同学你好,我想请问下,在《Long-Tailed Classifcation by Keeping the Good and Removing the Bad Momentum Causal Effect》这篇论文中,公式6以及对应的f(x,d;w)和g(x,d;w)都是怎么推导出来的?谢谢

请问de-counfound适用于其它optimizer吗?当我把代码里的optimizer换成adam后,结果很差,改变优化器,还有其它哪些配置需要做出相应的改变?

Thank you for your excellent work. I want to try to migrate the classifier (CausalNormClassifier) in this work to the task of multi label long tailel recognition. However, the following...

您好,感谢您分享的代码,我在训练模型的过程中出现了loss突然变成nan,acc变成0的问题,我分别从头开始进行了两次训练,但是还是产生了一样的问题。 我的训练环境是: - Nvidia RTX 2080Ti cuda10.1+cudnn7.6.3 - python3.8.5+pytorch1.5.1 - 数据集是ILSVRC2015 - 由于显存只有11G,所以我在config里把batch_size改成了64 - 其他不变 请问您知道可能的原因是什么吗?您用的ImageNet-LT是由ILSVRC2015提取的吗? ![loss2nan](https://user-images.githubusercontent.com/37700210/97077659-0db62880-1618-11eb-8da5-bdc38cb6f9e1.png) ![loss2nan2](https://user-images.githubusercontent.com/37700210/97077697-55d54b00-1618-11eb-9259-f4e6f70d9d8d.png)

**Describe the bug** training on COCO dataset is ok, but when I train on LVIS meet this bug. **Environment** 1. Please run `python mmdet/utils/collect_env.py` to collect necessary environment information and...