Three Tips for Boosting CNN Inference Performance (Download)
This article presents three tips that can help you significantly improve the performance of convolutional neural network (CNN) architectures for inference tasks. These tips are based on Recogni’s experience in successfully converting numerous neural networks (NNs) that have been trained on tasks ranging from simple image classification to more sophisticated challenges including semantic segmentation and 2D and 3D object detection. There’s real-world experience behind these tips.
Image classification is a computer vision application that uses a NN to recognize objects in a picture or a video frame by extracting and recognizing features in the image. A NN applied to semantic image segmentation attempts to label each pixel of an image with a corresponding object class.