weitianhan

@lemondan , thanks for your reply. In your Co-CNN paper, in the upsampling operations in the global-to-local stage, are you just doing bilinear upsampling without learning parameters in deconvolution layers?

Hi @lemondan, I would like to ask how do you determine the final output in the Co-cnn paper. For example the output of the network are C probability maps, then...

@lemondan , would you please give some hints about how do you do back-propagation in the superpixel layers?

Hi @hgaiser , I am basically trying the same thing as you, migrating this program to ResNet. And I am facing the same problem, Res-101 runs out of memory and...

Hi @hgaiser , thanks for your reply and contribution. I will look into it and have a try. By the way, are you feeding VOC dataset or COCO dataset?

OK, if the result is normal, it proves that the network makes sense. I will try it on COCO dataset.

Oh I didn't try the default setting. But after some modification I get 22% mAP in segmentation task using VGG16 which they report 19.5% in paper.

@zimenglan-sysu-512 What kind of configuration are you referring to?

Never mind the problem. I fix it by passing the library path in cmakelist. ```diff if (USE_RKNN) target_compile_definitions(piper PRIVATE USE_RKNN) target_sources(piper PRIVATE piper/rknn-inferer.cpp) - target_link_libraries(piper PRIVATE rknn) + target_link_libraries(piper PRIVATE...

Yes. I am developing app on an ARM development board. The wheel used to install python package is `pylance-0.10.18-cp39-abi3-manylinux_2_24_aarch64.whl`. But the same thing happens when I run this script on...