在转换模型时， 1_quantize_model.sh文件需要修改
--quantized-dtype dynamic_fixed_point-8
请问这一步是把模型的float型转换为int型吗？
改成
--quantized-dtype dynamic_fixed_point-16
-
这样能提高模型精度吗？
-
这样操作后，模型转换无误，但进行检测时，检测不到目标。
-
如何在转换模型时精度更高，比如不用int型，使用float型？
感谢告知
1 Like
Hey @penggeng, did you solve this issue? Can you please help me also in solving this issue?
嘿@penggeng，您解决了这个问题吗？您能帮我解决这个问题吗？
@CodeLogist I think @penggeng asked this in another post, check it there,
Quantification problem
2 Likes
no problem, just check the recent posts once again if you find a question that you think has not been addressed. if the problem is that serious than the person would have tried posting it again 
1 Like