Closed jveitchmichaelis closed 3 years ago
Sir, opencv 4.4 was released yesterday. It supports yolov4. I am going to install it.
Great, let me know if it works!
Sir, Opencv can run yolov4 and yolo-tiny4. But I use deeplabel to detect no bounding box appear.
I tried to re-install the deeplabel, it showed:
In file included from src/mainwindow.h:24:0, from src/mainwindow.cpp:1: src/detection/detectoropencv.h:11:9: fatal error: opencv2/dnn.hpp: No such file or directory
^~~~~~~~~~~~~~~~~
In file included from src/mainwindow.h:24:0, from src/main.cpp:1: src/detection/detectoropencv.h:11:9: fatal error: opencv2/dnn.hpp: No such file or directory
^~~~~~~~~~~~~~~~~
It may have moved in 4.4, try replacing with #include <opencv2/dnn/dnn.hpp>
It may have moved in 4.4, try replacing with
#include <opencv2/dnn/dnn.hpp>
Sorry. The reason of mentioned failure is I forgot to set path in Deeplabel.pro.
The file path is still opencv2/dnn/dnn.hpp on anaconda environment.
but make is still error:
/usr/bin/ld: cannot find -lopencv_tracking /usr/bin/ld: cannot find -lopencv_dnn collect2: error: ld returned 1 exit status Makefile:262: recipe for target 'release/DeepLabel' failed make: *** [release/DeepLabel] Error 1
Hmm ok so linker error - you're just missing the libs. Did you build those? (tracking requires contrib)
Hmm ok so linker error - you're just missing the libs. Did you build those? (tracking requires contrib)
Yes. You are correct. Now I go to fix it. Thank you.
@brianng0305 did it work? :)
@brianng0305 did it work? :)
Due to installed old version Deeplabel can run yolov4 so I gave up to install. Thanks.
As per this issue, it would be nice if we could support Yolov4, but currently we rely on opencv for backend processing. We could, alternatively, ask the user to provide a darknet binary/dll and use that.
This would also open the door for simple integration of training into the deeplabel GUI.
https://github.com/jveitchmichaelis/deeplabel/issues/17