LAVT-RIS issues

a mistake in class SpatialImageLanguageAttention

sim_map = sim_map + (1e4*l_mask - 1e4) # assign a very small number to padding positions I guess the 'small number' should be 1e-4?

Garfieldwwt

problem about downloading mscoco and saiapr_tc-12

1

the link of download mscoo and saiapr_tc-12 is down

SeptemberNightRain

About the network

作者您好，非常感谢您开源了您的优秀工作，我最近在做相关的研究，想请教这个工作的一些问题： 1. 作者您在某个issue提到如果batchsize太小的话网络训不起来，具体是性能下降多少？ 2. 如果backbone不采用预训练网络，而是随机初始化，是否也能使网络收敛？ 3. 对于损失函数的权重，您是如何决定为[0.9, 1.1]的呢？图片中有的物体过小，只靠权重可以克服吗？我是做其他领域的相关研究接触到了您的工作，对先前的工作和常用方法不太了解，还请您原谅并赐教！以下是上方文字的机翻：(The following is a machine translation of the text above) Hello author, thank you very much for opening up your excellent...

liuxuexun

issue with running code on colab

While executing the code on colab,i came across the following errors 1.Installing the requirements mentioned causes pip to break giving the error ``` [Error pip install with exit code: 1...

sparshgarg23

Misunderstanding about multi-class dice.

Thanks for your excllent work! I noticed that you mentioned the "multi-class dice loss" can take better performance. However, in my views, the final output masks and ground truths are...

Mr-Neko

About Dice Loss

3

Hello, author! Thank you for contributing an influential paper to the field of res. I have a small question to ask you regarding your statement on April 13th, 2023: "Using...

nini0919

demo

3

Randomly initialize Multi modal Swin Transformer weights.Traceback(most recent call last):File，line 64, indemo inference.py ,single bert model.load state dict(checkpoint[ bert model’])(eyError :bert mode

Joey666x