-
你好！我发现很多半监督语义分割的工作的classifier网络架构是不同的，比如ICLR2022的ReCO，它的分类器是把映射后的c1和经过ASPP的c4拼接起来，然后过两层卷积层，但是你们的分类器是过了三层卷积层。再如CVPR2022的U2PL，他们没有像你们和U2PL一样把c1的channel从256减到48，而是直接进行project，分类器卷积层的channel数就从304变成了512. …
-
Do you have any plan to implement the MAM method proposed in _(He et al., ICLR2022) TOWARDS A UNIFIED VIEW OF PARAMETER-EFFICIENT TRANSFER LEARNING_ ? It is an improvement on the unified level of the …
-
Hello.
In
`python3 -m domainbed.scripts.train\
--data_dir=/my/data/dir/\
--algorithm ERM\
--dataset OfficeHome\
--test_env ${test_env}\
--init_step\
…
-
Dear SQuant team,
I am a college student of computer science. I found your paper accepted by ICLR2022 named SQuant: On-the-fly Data-free Quantization Via Diagonal Hessian Approximation. I am deeply…
-
微博内容精选
-
Good job, bro. I would like to share our recent work [Towards better understanding and better generalization of few-shot classification in histology images with contrastive learning] (https://arxiv.or…
yjump updated
2 years ago
-
Great work - results look very impressive.
From my experience Diffusion Models are quite heavy in terms of computation, can you share info about the inference speed?
-
您好！我最近在了解resnet的ann2snn的工作，我跑了spikingjelly中的示例，取得了不错的效果，但是我对其内部工作原理还不是很清楚。示例中用到的是最大激活归一化，但是对shortcut是如何处理的呢？十分期待您的解答！
-
您好！
感谢你们的努力，创造出这个非常有价值的工具！
请问能否加入[ICLR](https://iclr.cc/)的截止时间呢？虽然ICLR暂时未被列入CCF，但在国内外机器学习领域都有很大的影响力，国内很多高校的评定中，ICLR都是等同于CCF A的存在。
盼回复，谢谢！
-
I tried replicating the results in your paper following your instructions, however it seems like the parameters are not correct. For example, when I run:
`poetry run topognn/train_model.py --model To…