└── README.md /README.md: -------------------------------------------------------------------------------- 1 | # CNN-heatmap 2 | 3 | ~~~ 4 | 5 | 1. Visualizing and Understanding Convolutional Networks 6 | https://arxiv.org/pdf/1311.2901v3.pdf 7 | https://neukom.dartmouth.edu/docs/bbat-wacv2016.pdf 8 | http://cs231n.github.io/understanding-cnn/ 9 | 10 | 2. Net surgery trick 11 | http://cs231n.github.io/convolutional-networks/#convert 12 | https://github.com/BVLC/caffe/blob/master/examples/net_surgery.ipynb 13 | 14 | https://leonardoaraujosantos.gitbooks.io/artificial-inteligence/content/image_segmentation.html 15 | https://arxiv.org/pdf/1502.02766v3.pdf 16 | https://arxiv.org/pdf/1411.4038v2.pdf 17 | http://www.cv-foundation.org/openaccess/content_cvpr_2015/papers/Long_Fully_Convolutional_Networks_2015_CVPR_paper.pdf 18 | https://arxiv.org/pdf/1605.06211v1.pdf 19 | 20 | `convert fully connected layers to their 21 | equivalent convolutional layers, since the weights are the same and only 22 | the shapes are different.` 23 | 24 | 3. Global average pooling layer 25 | Network In Network 26 | https://arxiv.org/pdf/1312.4400.pdf 27 | 28 | 4. Learning Deep Features for Discriminative Localization 29 | http://cnnlocalization.csail.mit.edu/ 30 | https://github.com/jacobgil/keras-cam 31 | 32 | 5. Grad-CAM: Gradient-weighted Class Activation Mapping 33 | https://github.com/ramprs/grad-cam 34 | http://gradcam.cloudcv.org/ 35 | https://arxiv.org/pdf/1610.02391v2.pdf 36 | 37 | 6. Is object localization for free? – Weakly Supervised Object Recognition with Convolutional Neural Networks 38 | http://www.di.ens.fr/willow/research/weakcnn/ 39 | http://www.di.ens.fr/willow/research/cnn/ 40 | 41 | ~~~ 42 | 43 | Additional: 44 | ~~~ 45 | 46 | 1. Deep Inside Convolutional Networks: Visualising Image Classification Models and Saliency Maps 47 | https://arxiv.org/pdf/1312.6034v2.pdf 48 | 49 | 2. Top-down NeuralAttention by Excitation Backprop (c-MWP) 50 | https://arxiv.org/pdf/1608.00507v1.pdf 51 | http://cs-people.bu.edu/jmzhang/excitationbp.html 52 | 53 | 3. Paying More Attention to Attention: Improving the Performance of Convolutional Neural Networks via Attention Transfer 54 | https://arxiv.org/pdf/1612.03928v1.pdf 55 | https://github.com/szagoruyko/attention-transfer 56 | 57 | 4. Visualizing Deep Neural Network Decisions: Prediction Difference Analysis 58 | https://openreview.net/pdf?id=BJ5UeU9xx 59 | https://github.com/lmzintgraf/DeepVis-PredDiff 60 | A New Method to Visualize Deep Neural Networks 61 | https://icmlviz.github.io/assets/papers/23.pdf 62 | 63 | 5. Self-Taught Object Localization with Deep Network 64 | https://arxiv.org/pdf/1409.3964.pdf 65 | 66 | 6. Shallow and Deep Convolutional Networks for Saliency Prediction 67 | https://github.com/imatge-upc/saliency-2016-cvpr 68 | ~~~ 69 | 70 | Keras codebase: 71 | ~~~ 72 | https://github.com/raghakot/keras-vis 73 | 74 | Oclusion based technique: 75 | https://github.com/waleedka/cnn-visualization/blob/master/cnn_visualization.ipynb 76 | 77 | Net surgery trick: 78 | https://github.com/heuritech/convnets-keras 79 | 80 | GAP-CAM 81 | https://github.com/alexisbcook/ResNetCAM-keras/ 82 | https://github.com/jacobgil/keras-cam 83 | https://github.com/tdeboissiere/VGG16CAM-keras 84 | https://github.com/keras-team/keras/blob/0cfa5c2709906a7a76f552f71a562f899e408695/examples/class_activation_maps.py 85 | 86 | Grad-CAM 87 | https://github.com/jacobgil/keras-grad-cam 88 | https://github.com/hiveml/tensorflow-grad-cam 89 | 90 | https://raghakot.github.io/keras-vis/visualizations/attention/ 91 | 92 | https://github.com/mlhy/ResNet-50-for-Cats.Vs.Dogs 93 | ~~~ 94 | 95 | Tensorflow codebase: 96 | ~~~ 97 | GAP 98 | https://github.com/sjchoi86/tensorflow-101/blob/master/notebooks/gap_mnist.ipynb 99 | 100 | GAP-CAM 101 | https://github.com/jazzsaxmafia/Weakly_detector 102 | 103 | Grad-CAM 104 | https://github.com/Ankush96/grad-cam.tensorflow 105 | 106 | ~~~ 107 | 108 | PyTorch codebase: 109 | ~~~ 110 | Grad-CAM 111 | https://github.com/kazuto1011/grad-cam-pytorch 112 | https://github.com/jacobgil/pytorch-grad-cam 113 | 114 | GAP-CAM 115 | https://github.com/metalbubble/CAM/blob/master/pytorch_CAM.py 116 | 117 | SPN 118 | https://github.com/yeezhu/SPN.pytorch 119 | 120 | https://github.com/utkuozbulak/pytorch-cnn-visualizations 121 | 122 | https://github.com/jacobgil/pytorch-explain-black-box 123 | ~~~ 124 | 125 | Regression Activation Map 126 | https://github.com/cauchyturing/kaggle_diabetic_RAM 127 | 128 | 129 | Other 130 | ~~~ 131 | https://github.com/metalbubble/cnnvisualizer 132 | 133 | https://github.com/InFoCusp/tf_cnnvis 134 | 135 | https://github.com/ppwwyyxx/tensorpack/tree/master/examples/Saliency 136 | 137 | https://github.com/InFoCusp/tf_cnnvis 138 | 139 | https://github.com/keplr-io/quiver 140 | 141 | https://jacobgil.github.io/deeplearning/vehicle-steering-angle-visualizations 142 | https://jacobgil.github.io/deeplearning/class-activation-maps 143 | 144 | https://jacobgil.github.io/deeplearning/filter-visualizations 145 | https://jacobgil.github.io/computervision/saliency-from-backproj 146 | 147 | https://github.com/CSAILVision/NetDissect 148 | 149 | https://medium.com/merantix/picasso-a-free-open-source-visualizer-for-cnns-d8ed3a35cfc5 150 | 151 | https://github.com/CSAILVision/NetDissect 152 | http://netdissect.csail.mit.edu/ 153 | 154 | https://github.com/imatge-upc/saliency-salgan-2017 155 | 156 | http://imatge-upc.github.io/saliency-2016-cvpr/ 157 | https://github.com/imatge-upc/saliency-2016-cvpr 158 | 159 | https://openreview.net/pdf?id=SkfMWhAqYQ 160 | 161 | https://thegradient.pub/a-visual-history-of-interpretation-for-image-recognition/ 162 | https://thegradient.pub/interpretability-in-ml-a-broad-overview/ 163 | 164 | ~~~ 165 | 166 | Visualization: 167 | ~~~ 168 | https://github.com/shaohua0116/Activation-Visualization-Histogram 169 | ~~~ 170 | 171 | To look at: 172 | ~~~ 173 | http://blog.qure.ai/notes/visualizing_deep_learning 174 | https://github.com/utkuozbulak/pytorch-cnn-visualizations 175 | https://github.com/fornaxai/receptivefield 176 | ~~~ 177 | --------------------------------------------------------------------------------