Description
A CoordConv layer is a simple extension to the standard convolutional layer. It has the same functional signature as a convolutional layer, but accomplishes the mapping by first concatenating extra channels to the incoming representation. These channels contain hard-coded coordinates, the most basic version of which is one channel for the coordinate and one for the coordinate.
The CoordConv layer keeps the properties of few parameters and efficient computation from convolutions, but allows the network to learn to keep or to discard translation invariance as is needed for the task being learned. This is useful for coordinate transform based tasks where regular convolutions can fail.
Papers Using This Method
Mammographic Breast Positioning Assessment via Deep Learning2024-07-15Semi-Supervised Domain Adaptation for Wildfire Detection2024-04-02YOLOrtho -- A Unified Framework for Teeth Enumeration and Dental Disease Detection2023-08-11PP-YOLOE: An evolved version of YOLO2022-03-30In Defense of Kalman Filtering for Polyp Tracking from Colonoscopy Videos2022-01-27Depth-aware Object Segmentation and Grasp Detection for Robotic Picking Tasks2021-11-22Segmentation of Drilled Holes in Texture Wooden Furniture Panels Using Deep Neural Network2021-05-23PP-YOLOv2: A Practical Object Detector2021-04-21An Efficient Multitask Neural Network for Face Alignment, Head Pose Estimation and Face Tracking2021-03-13PP-YOLO: An Effective and Efficient Implementation of Object Detector2020-07-23PropagationNet: Propagate Points to Curve to Learn Structure Information2020-06-25Deep feature fusion for self-supervised monocular depth prediction2020-05-16Segmentation-based Method combined with Dynamic Programming for Brain Midline Delineation2020-02-27Adaptive Wing Loss for Robust Face Alignment via Heatmap Regression2019-04-16Automatic salt deposits segmentation: A deep learning approach2018-11-21An Intriguing Failing of Convolutional Neural Networks and the CoordConv Solution2018-07-09