ChenYingpeng / caffe-yolov3

A real-time object detection framework of Yolov3/v4 based on caffe
473 stars 231 forks source link

how to convert yolo layer to caffe #7

Open fzhu129 opened 6 years ago

fzhu129 commented 6 years ago

Hi Mr. Chen:

When converting yolov3 to caffe proto, there is warning for yolo level not available?

Could you share the code upon yolo_layer.hpp yolo_layer.cpp if possible? or could you give us more detail upon how to add yolo layer into caffe?

Thanks you very much for your kindness in advance.

Wei

ChenYingpeng commented 6 years ago

Sorry,I don't write yolo layer into caffe , and just rewrite it with cpp code.If you want to write it into caffe ,you should write it by youeself.

fzhu129 commented 6 years ago

Hi Mr.Chen: I can create the same yolov3.proto but yolo layer missed. That is why some layer lost.

Thank you very much for your sharing your effort with us. Wei

fzhu129 commented 6 years ago

Hi Mr.Chen:

Which caffe version used by you, I have added yolo layer into caffe with cpu version.

but I got this:

F0818 16:54:27.777938 16953 blob.hpp:140] Check failed: num_axes() <= 4 (773770 vs. 4) Cannot use legacy accessors on Blobs with > 4 axes. Check failure stack trace:

I don't use jetson TX2 so I just use the normal ubuntu with cpu.

Could you tell me what version of caffe you used?

Thanks Wei

ChenYingpeng commented 5 years ago

I rewrited it with gpu.So if you wanted to inference just using cpu, you should rewrite it followed darknet.

fzhu129 commented 5 years ago

problem fixed. Thank you very much