1x1 Convolution
التفاف 1×1
التفاف بنواة مقاسها 1×1 يعمل عبر القنوات فقط دون سياق مكاني. يُستخدم لتقليص عدد القنوات (عنق الزجاجة) قبل الالتفافات المُكلفة ولإضافة تحويل لاخطي عبر تنشيط ReLU.
A convolution with a 1×1 kernel that operates across channels only, without spatial context. Used for channel dimensionality reduction (bottleneck) before expensive convolutions and to add nonlinearity via ReLU activation.
Also translated asالتفاف نقطي، الشبكة داخل الشبكة
First appears in this corpus in: Going Deeper with Convolutions (2014)
Appears in these papers
- A ConvNet for the 2020s2022in the sky ✦
- A ConvNet for the 2020s2022in the sky ✦
- EfficientNet: Rethinking Model Scaling for Convolutional Neural Networks2019in the sky ✦
- Going Deeper with Convolutions2014in the sky ✦
- MobileNets: Efficient Convolutional Neural Networks for Mobile Vision Applications2017in the sky ✦
- Pixel Recurrent Neural Networks2016in the sky ✦
- SegFormer: Simple and Efficient Design for Semantic Segmentation with Transformers2021in the sky ✦
- SegFormer: Simple and Efficient Design for Semantic Segmentation with Transformers2021in the sky ✦
- YOLOv3: An Incremental Improvement2018in the sky ✦