keyword
Deformable CNN
A deformable CNN, or deformable convolutional neural network, is a deep learning architecture designed to improve visual recognition by dynamically adjusting its sampling locations to accommodate geometric transformations. Unlike conventional convolutional neural networks that sample feature maps using rigid, regular grids, a deformable CNN introduces learnable spatial offsets to standard convolutional and pooling operations. These offsets are predicted directly from the input features during training without additional supervision, allowing the receptive field to freely deform and align with the actual scale, pose, and boundary of non-rigid objects. By enabling adaptive spatial sampling, deformable CNNs enhance feature extraction for complex computer vision tasks such as object detection, instance segmentation, and pose estimation.
1 item
