Bing Han comments

Results 60 comments of


                                            Bing Han

NLLB vocabulary missing common Chinese character/tokens

Other character base languages, such as Japanese, may have similar problems too.

不知道怎么形容，但是NLLB 200 demo所展示的，翻译为简体中文的效果真的太差劲了。

之前M2M100我也试用过，连最基本的英译中这种语料很丰富的语种都翻译的很不好。想问下NLLB的demo演示在哪里可以看到呢？

如何引用Installation 中的库

感谢issue 文件中有这样一句，应该可以将头文件供他人使用 ``` set_target_properties(mymath PROPERTIES PUBLIC_HEADER ${CMAKE_SOURCE_DIR}/include/mymath.h) ```

GPU-based NMS method was used to predict the results, and multiple calibration boxes appeared on each face.

Can you show me the input image and the output result?

Training from scratch

Yes，check the doc folder please.

confusion regarding map output feature to the location of original image

I think it's my fault. You can remove the "+1". But it's ok if calculate the same formulation in both training and evaluating process. And also the 1 pixel error...

How much speed on GPU

It depends on image size and how many faces in the images.

Is it better to handle for the case num = int(line.strip()) = 0?

Is there any case that there is no bbox in a image? If so, I will handle this case.

Is it better to handle for the case num = int(line.strip()) = 0?

Could you please send me a pull request to handle this problem?

error while detect image

This project was only tested on python3. Is there the same error on python3?