arxiv: https://arxiv.org/abs/1803.01534
this paper propose a network structure(“PANet”) with bottom up path augmentation and a few other tweak methods which results in better instance segmentation performance.
(more…)arxiv: https://arxiv.org/pdf/1807.06521.pdf
trying to attach attention module to CNN. but instead of blindly attaching it which would compute a 3D attention map which is computationally expensive, this work proposes to compute spatial attention and channel attention separately which achieves a similar effect with much less parameters.
(more…)arxiv: https://arxiv.org/abs/1406.4729
pools a fixed number of features from final feature map of backbone which can be fed to the dense layer afterwards, thus allowing network to work in a non-fixed input size manner.
(more…)arxiv: https://arxiv.org/abs/2003.12294
this approach will try to take global context into account by adopting transformer, and also use similar structure to reason semantic content.
(more…)arxiv: https://arxiv.org/pdf/1704.04861.pdf key points focus on optimizing for latency, small networks. use depthwise separable convolutions, to reduce computation as much as possible further reduce size of models based on width/resolution multiplier, but at the cost of accuracy depthwise separable convolution Read more…
arxiv: https://arxiv.org/abs/1903.03273
arxiv link: https://arxiv.org/abs/1709.08429
arxiv: https://arxiv.org/abs/2005.10821
arxiv link: https://arxiv.org/abs/1911.09070 key points multi scale with weighted bi-directional fpn. model scaling. compound scaling method, which jointly scales up resolution/depth/width for all backbone, feature network, box/class prediction network. use efficientnet backbone Bi-directional FPN Here are the key points of Read more…
good article from nvidia: https://developer.nvidia.com/blog/cuda-refresher-cuda-programming-model/ when in comes to size of concept: grid > block > thread max thread in one block: 1024