New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Inference #11
Comments
The Tensorrt softmax layer api dose not support axis:2. You must implement it by yourself. |
Hi @chenzhi1992. I can not understand in what format the data to the enqueue function come to me. I realized that the data comes in an enqueue function in the form of a vector in which (for 21 classes of pascal voc) there are 12764 elements located in the vector sequentially. That is, in the function as an input parameter there is a pointer to a vector of 268044 elements.
The kernel source code is simple, and I will not post it.
After these conversions, it does not look like I see what I need at the output.
|
What is the ConcatPlugin class implemented for? This class is not used anywhere. Instead it is used createConcatPlugin declared NvInferPlugin.h.
|
|
@chenzhi1992 , I've reviewed the code caffe softmax layer https://github.com/BVLC/caffe/blob/master/src/caffe/layers/softmax_layer.cu. The main problem for me is how to set the axis for maximum summation and exponent operations. I understand correctly, I get a pointer to a vector of 1x268044? How is the axis defined in this case? |
Hi @Maxfashko, I am also stuck with the same problem. Were you able to fix it, and if so could you tell how? |
@tianfanzhu, no, I did not solve this problem. I have other tasks, but as soon as I have time, I will study the code caffe |
@Maxfashko Can you update your kernel functions for SoftmaxLayer here? |
@linux-devil, I hope I will help more than the author of this repository
|
@Maxfashko how did you get label name from label.txt? Thanks |
@quocbh https://github.com/Maxfashko/NV_TRT_SSD enjoy yourself |
In void cudaSoftmax(int n, int channels, float* x, float*y) , what is n? how you calculate n? |
should add one line after "shared float sum_value;"
|
Hey. I implemented the inference using this code. But it seems that the boxes do not correspond to the actual location of objects in the image. I did not implement the softmax layer in the code, instead I used the usual softmаx caffe layer.
Is it necessary to use the softmax layer, implemented by myself?
The text was updated successfully, but these errors were encountered: