Closed spinoza1791 closed 6 years ago
Yes, I will try it.
@spinoza1791 I added branch yolov3. But it is not work (killed with segmentation fault). I think Raspberry Pi 3's memory is not sufficient to run YOLOv3. It runs well on PC with NNPACK.
Is there any solution to make yolov3 run on Rpi3?
do you have yolov3 NNPack Darknet working for any CPU? Do you think the seg fault is only for Pi?
On Tue, Mar 27, 2018, 8:58 AM Thomas Park notifications@github.com wrote:
@spinoza1791 https://github.com/spinoza1791 I added branch yolov3. But it is not work (killed with segmentation fault). I think Raspberry Pi 3's memory is not sufficient to run YOLOv3.
— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/digitalbrain79/darknet-nnpack/issues/19#issuecomment-376535187, or mute the thread https://github.com/notifications/unsubscribe-auth/AZX8mAau2VXFIl2xqf20ak33fZVmi_L0ks5tikWJgaJpZM4S6NtZ .
@spinoza1791 yolov3 run well on Intel I7 CPU. If you want run yolov3 on Intel CPU, you should set ARM_NEON=0 in makefile. I did not test other CPUs except Intel I7 and RPI3.
There are new files in current darknet which are not in your repo. Such as znorm_layer.c/.h, logistic_layer.c/h, unsample_layer.c/h and yolo_layer.c/h. You will need to include these to take advantage of new yolo features. Will you be updating NNPack/darknet to include these?
when using the upsample layer i get "type not recognized: [upsample]' How is are you getting NNPack to work without the unsample_layer.c/h in your repo?
@spinoza1791 Did you change branch to yolov3? (git checkout yolov3) I cannot find znorm_layer.c/.h and unsample_layer.c/h in original darknet repo.
Here's a link to the current yolo repo, updated 27 days ago for the upsample.c https://github.com/pjreddie/darknet/tree/master/src
'git checkout yolov3' works with upsample:) Thank you! You may close issue.
If you write a yolov3 from AlexeyAB darkent other than pjreddie darknet ,it will works.The fatal error origins from pjreddie darknet.
Will you be updating darknet-nnpack to make use of new features in YOLOv3 today? https://github.com/pjreddie/darknet