Tag:network structure

  • Application of automatic web search (NAS) in semantic segmentation (2)

    Time:2021-9-16

    preface:This paper will introduce how to search the semantic segmentation model based on proxylessnas. The final search model structure can reach 36 FPS test results on the CPU, and show the application of automatic network search (NAS) in semantic segmentation. With the advent of neural architecture search technology, deep learning has gradually developed to the […]

  • Application of automatic network search (NAS) in semantic segmentation (I)

    Time:2021-8-28

    【abstract】This paper briefly introduces the development status of NAS and its application in semantic segmentation, and interprets two popular works: darts and auto deeplab in detail. Automatic web search Most neural network structures are based on some mature backbones, such as RESNET and mobilenet, which are slightly improved to complete different tasks. Because of this, […]

  • Darts: classic network search method based on gradient descent, open end-to-end network search | ICLR 2019

    Time:2021-6-11

    Darts is a classic NAS method, which breaks the previous discrete network search mode and can carry out end-to-end network search. Because darts updates the network based on gradient, the direction of update is more accurate, and the search time is greatly improved compared with the previous method. Cifar-10 only needs 4gpu days.  Source: Xiaofei’s […]

  • Fbnet / fbnetv2 / fbnetv3: Facebook’s lightweight network exploration in NAS field | lightweight network

    Time:2021-6-7

    Fbnet series is a lightweight network series completely based on NAS method. The shortcomings of current search methods are analyzed and innovative improvements are gradually added. Fbnet combines DNAs and resource constraints, fbnetv2 adds channel and input resolution search, and fbnetv3 uses accuracy prediction for fast network structure search  Source: Xiaofei’s algorithm Engineering Notes official […]

  • Handwritten numeral image recognition convolution neural network

    Time:2021-4-24

    Import dependency from tensorflow import keras from matplotlib import pyplot as plt from tensorflow.keras.layers import Conv2D, MaxPool2D, Flatten, Dense   Download dataset MNIST data set is a public handwritten numeral data set. There are 7W 28 * 28 pixels of 0-9 handwritten numeral images and labels, of which 6W are training sets and 1W are […]

  • A CNN example of torch convolutional neural network

    Time:2021-4-2

    Training of torch convolution neural network I won’t say more about the basic knowledge of convolutional neural network (CNN) here. For detailed information, please refer to my explanation in CSDN Principle and process of CNN convolutional neural network The following is a visual display of convolution process website https://www.cs.ryerson.ca/~aharley/vis/conv/ 1、 Using Python to train MINST […]

  • Training of cifar10 dataset

    Time:2021-3-25

    Download dataset The cifar10 dataset contains 60000 32 * 32 pixel color images and tags, covering ten categories: aircraft, automobile, bird, cat, deer, dog, frog, horse, boat and truck. Fifty thousand of them were for training and ten thousand for testing.   import tensorflow as tf from tensorflow import keras from matplotlib import pyplot as […]

  • Paper reading: u-net

    Time:2020-11-16

    introduction This paper introduces a U-shaped network structure for semantic segmentation. In fact, it is based on the idea of encoding and decoding, which can effectively combine the information of low resolution and high resolution, and can better segment the image edge. It was proposed in the same year as FCN, but u-net uses a […]

  • Paper reading: eNet

    Time:2020-10-30

    introduction Pixel level classification tasks such as semantic segmentation have been developed a lot, and the existing models are developing towards higher and higher accuracy. However, such tasks also have important applications for embedded devices, so real-time semantic segmentation is essential. However, most of the models are towards high-precision, which is very poor in real-time. […]

  • Thesis reading: eNet

    Time:2020-6-4

    introduction Pixel level classification tasks such as semantic segmentation have developed a lot, and the existing models are moving towards higher and higher precision. But such tasks also have important applications for embedded devices, so it is necessary to achieve real-time semantic segmentation, but most of the models are towards high precision, which is very […]

  • How to write network code (keras) according to network diagram

    Time:2020-5-30

    How to write network code (keras) according to network diagram Sometimes, for some purpose, we want to build our own network step by step, rather than simply call the existing model, The package is so powerful that I almost forget the internal structure I have a small idea to add to the network structure, in […]

  • MASK-RCNN(2)

    Time:2020-4-21

    2. Network Architecture The network is divided into two parts: the first part is the backbone convolution network, which is used to extract the features of the whole image; the second part is the head, which is used to process ROI, which is divided into two branches, one is used to classify and regress box, the […]