About testing result about rstnet HOT 16 CLOSED

你好，能更新代码吗，这个精度确实不对。

from rstnet.

I use the pretrained model provide in this project to test COCO test set, and eval performance is 'BLEU': [0.806718016311047, 0.6540980786273354, 0.5092295283787919, 0.3898375465182736], 'METEOR': 0.29228558925065895, 'ROUGE': 0.585628568332276, 'CIDEr': 1.3206302796307086.
It seems different from the performance claimed in paper.
Could you provide the inference captions of COCO test set by using the method proposed in this paper.

from rstnet.

你好，能否提供一份特征？感激不尽！！

from rstnet.

from rstnet.

I use the pretrained model provide in this project to test COCO test set, and eval performance is 'BLEU': [0.806718016311047, 0.6540980786273354, 0.5092295283787919, 0.3898375465182736], 'METEOR': 0.29228558925065895, 'ROUGE': 0.585628568332276, 'CIDEr': 1.3206302796307086. It seems different from the performance claimed in paper. Could you provide the inference captions of COCO test set by using the method proposed in this paper.

关于language model，您有没有使用我上传的的模型参数文件？还是您重新训练的？

from rstnet.

I use the pretrained model provide in this project to test COCO test set, and eval performance is 'BLEU': [0.806718016311047, 0.6540980786273354, 0.5092295283787919, 0.3898375465182736], 'METEOR': 0.29228558925065895, 'ROUGE': 0.585628568332276, 'CIDEr': 1.3206302796307086. It seems different from the performance claimed in paper. Could you provide the inference captions of COCO test set by using the method proposed in this paper.

关于language model，您有没有使用我上传的的模型参数文件？还是您重新训练的？

我使用的是上传的的模型参数文件

from rstnet.

decoder = TransformerDecoderLayer(len(text_field.vocab), 54, 3, text_field.vocab.stoi['<pad>'], 'Pre-trained_Models/language_context.pth')

from rstnet.

I use the pretrained model provide in this project to test COCO test set, and eval performance is 'BLEU': [0.806718016311047, 0.6540980786273354, 0.5092295283787919, 0.3898375465182736], 'METEOR': 0.29228558925065895, 'ROUGE': 0.585628568332276, 'CIDEr': 1.3206302796307086. It seems different from the performance claimed in paper. Could you provide the inference captions of COCO test set by using the method proposed in this paper.

关于language model，您有没有使用我上传的的模型参数文件？还是您重新训练的？

我使用的是上传的的模型参数文件

我check一下。

from rstnet.

I use the pretrained model provide in this project to test COCO test set, and eval performance is 'BLEU': [0.806718016311047, 0.6540980786273354, 0.5092295283787919, 0.3898375465182736], 'METEOR': 0.29228558925065895, 'ROUGE': 0.585628568332276, 'CIDEr': 1.3206302796307086. It seems different from the performance claimed in paper. Could you provide the inference captions of COCO test set by using the method proposed in this paper.

关于language model，您有没有使用我上传的的模型参数文件？还是您重新训练的？

我使用的是上传的的模型参数文件

我check一下。

请问有结果了吗

from rstnet.

I use the pretrained model provide in this project to test COCO test set, and eval performance is 'BLEU': [0.806718016311047, 0.6540980786273354, 0.5092295283787919, 0.3898375465182736], 'METEOR': 0.29228558925065895, 'ROUGE': 0.585628568332276, 'CIDEr': 1.3206302796307086. It seems different from the performance claimed in paper. Could you provide the inference captions of COCO test set by using the method proposed in this paper.

关于language model，您有没有使用我上传的的模型参数文件？还是您重新训练的？

我使用的是上传的的模型参数文件

我check一下。

请问有结果了吗

您好，我前几天也测试了一下，确实是您说的那个结果，应该是这个Project的bug，抱歉。下面是我使用原始Project测试的结果：

这段时间在宿舍隔离，连接不上我们实验室的服务器了，就专心写毕业论文了。隔离结束后再更新这个Project。

from rstnet.

I use the pretrained model provide in this project to test COCO test set, and eval performance is 'BLEU': [0.806718016311047, 0.6540980786273354, 0.5092295283787919, 0.3898375465182736], 'METEOR': 0.29228558925065895, 'ROUGE': 0.585628568332276, 'CIDEr': 1.3206302796307086. It seems different from the performance claimed in paper. Could you provide the inference captions of COCO test set by using the method proposed in this paper.

Hello Wan！I also use this open source code to train the RSTNet. However, at the end of cross-entropy training stage, I only get the CIDEr score about 103, is that normal? How much CIDEr score did you get at the end of cross-entropy training stage? How many epochs did you train during RL stage? Did you use other tricks that can tell me?
Thank you very much! I expect your reply.

from rstnet.

Hello Wan！I also use this open source code to train the RSTNet. However, at the end of cross-entropy training stage, I only get the CIDEr score about 103, is that normal? How much CIDEr score did you get at the end of cross-entropy training stage? How many epochs did you train during RL stage? Did you use other tricks that can tell me?
Thank you very much! I expect your reply.

I do not train the RSTNet.

from rstnet.

你好，能更新代码吗，这个精度确实不对。
是啊，这个模型跑的结果比不上论文里面的，我不知道哪出了问题，作者能帮解答一下吗？谢谢了

已更新

from rstnet.

代码已更新，下面是测试结果截图。

from rstnet.

I use the pretrained model provide in this project to test COCO test set, and eval performance is 'BLEU': [0.806718016311047, 0.6540980786273354, 0.5092295283787919, 0.3898375465182736], 'METEOR': 0.29228558925065895, 'ROUGE': 0.585628568332276, 'CIDEr': 1.3206302796307086. It seems different from the performance claimed in paper. Could you provide the inference captions of COCO test set by using the method proposed in this paper.

关于language model，您有没有使用我上传的的模型参数文件？还是您重新训练的？

我使用的是上传的的模型参数文件

我check一下。

请问有结果了吗

已更新

from rstnet.

I use the pretrained model provide in this project to test COCO test set, and eval performance is 'BLEU': [0.806718016311047, 0.6540980786273354, 0.5092295283787919, 0.3898375465182736], 'METEOR': 0.29228558925065895, 'ROUGE': 0.585628568332276, 'CIDEr': 1.3206302796307086. It seems different from the performance claimed in paper. Could you provide the inference captions of COCO test set by using the method proposed in this paper.

关于language model，您有没有使用我上传的的模型参数文件？还是您重新训练的？

我使用的是上传的的模型参数文件

我check一下。

请问有结果了吗

已更新

谢谢更新

from rstnet.