Learning Deep Features for Discriminative Localization

Posted 2022-12-06 zbxzc

Learning Deep Features for Discriminative Localization论文笔记以及Caffe实现

论文笔记 | Learning Deep Features for Discriminative Localization

论文笔记： Learning Deep Features for Discriminative Localization

可以利用论文的思路对原始以224*224或227*227尺寸图像作为输入的AlexNet或VGGNet使用更高分辨率的图像作为输入，主要是为了利用原来在imagenet上预训练的模型进行finetuning。

https://github.com/hshota0530/caffe_models

VGG的论文里还有提到用384x384，512x512的网络进行训练，然后再融合模型来提高精度，增加384x384的输入，类似于256x256，网络的输入参数大小为336x336，和256x256一样，随机截取作为输入，第一个卷积层把stride改为3就可以。512x512的网络由于输入图片太大，只做了两次实验，一次是卷积大小依旧7x7，stride改为4，不成功。一次是加入两个5x5，stride为2的卷积层，依旧不行，然后就没再实验。 

https://github.com/metalbubble/CAM

https://github.com/jacobgil/keras-cam

https://github.com/slundqui/DeepGAP

https://github.com/markdtw/class-activation-mapping

https://github.com/jacobgil/pytorch-grad-cam

https://github.com/jacobgil/keras-grad-cam

https://github.com/taoyilee/Keras_MedicalImgAI/blob/master/app/grad_cam.py

https://github.com/insikk/Grad-CAM-tensorflow

https://github.com/adityac94/Grad_CAM_plus_plus

https://github.com/conan7882/CNN-Visualization

Visualizing and Understanding Convolutional Networks 阅读笔记-网络可视化NO.1

Grad-CAM: Visual Explanations from Deep Networks via Gradient-Based Localization

凭什么相信你，我的CNN模型？（篇一：CAM和Grad-CAM)

论文笔记：WILDCAT: Weakly Supervised Learning of Deep ConvNets

Weakly supervised Localization using deep feature maps

论文笔记： Weakly Supervised Semantic Segmentation Using Superpixel Pooling Network