-
https://bindog.github.io/blog/2019/10/23/why-triplet-loss-works/
0x00 triplet loss简介0x00 triplet loss简介深度学习领域有一块非常重要的方向称之为metric learning，其中一个具有代表性的方法就是triplet loss，triplet loss的基本思想很清晰，就是让同一类别样本…
-
I observed the training loss (for both long_los and icu_mortality tasks in the [MIMIC IV tutorial](https://meds-torch.readthedocs.io/en/latest/tutorial/)) is unstable and nonconverging for supervised …
-
首先非常感谢您提供的代码，有两个问题希望您能提供帮助
1、在我用batch_hard_triplet_loss方法训练手写数字网络时，我发现如果我的网络不加BN层,训练一会之后loss就会固定在margin的大小不在下降，并且acc（acc使用KNN求得）变成了0，我查看了一下_pairwise_distances，发现距离矩阵也差不多都变成了？这是什么原因造成的呢，必须要加bn层吗？
2、我…
-
Hi!,
I am trying to do fine-tuning of the last layer of the arcface model, for this leaving frozen all the weights of the network, except those belonging to the last layers. The problem is that the…
-
tensor([0.8285, 0.7775, 0.6481, 0.6993, 0.6650, 0.7633, 0.7279, 0.6936, 0.6447,
…
-
Thank you very much for your excellent work.
One problem I am confused about is the definition of the `crossmodal loss function` and` coseparation loss function`. In the train.py, why random numbers …
-
In version 1.2.0, Centroid Triplet Loss is unstable. When the batch size is small, the below error always occurs. If larger batch size, it is relatively be relieved.
`File "/root/miniconda3/lib/pyt…
-
I had try to train the model by myself.
I download CASIA-WebFace which had been celan and then follow the instructions in wiki "Triplet loss training" to train the mode.
But it seems the accuracy …
-
你好作者，这是一个非常棒的工作，我想和您请教沟通下如何对模型做轻量化的部署。
1. 降低ResConv层数及channel。
2. 对上下采样做更快的，例如使用conv替换。
3. 估计1/2，1/4甚至1/8的光流，上采样至原分辨率做warp
4. grid_sample操作较为耗时，是否有更好的替换方案？可不可以不内建warp
5. 为了解决4，是否可以在不做warp的情况下做多…
-
Hello, I am a Korean student who loves the work of sentence-transformers team.
I've used a lot of losses you made when finetuning an embedding model.
However, I thought that it will be nice if the…