zhoubolei / Cnnvisualizer
Licence: mit
Visualizer for Deep Neural Networks
Stars: ✭ 259
Programming Languages
c
50402 projects - #5 most used programming language
Labels
Projects that are alternatives of or similar to Cnnvisualizer
Dfl Cnn
This is a pytorch re-implementation of Learning a Discriminative Filter Bank Within a CNN for Fine-Grained Recognition
Stars: ✭ 245 (-5.41%)
Mutual labels: cnn
Audio-Classification-using-CNN-MLP
Multi class audio classification using Deep Learning (MLP, CNN): The objective of this project is to build a multi class classifier to identify sound of a bee, cricket or noise.
Stars: ✭ 36 (-86.1%)
Mutual labels: cnn
RCAN-tf
TensorFlow code for ECCV 2018 paper "Image Super-Resolution Using Very Deep Residual Channel Attention Networks"
Stars: ✭ 25 (-90.35%)
Mutual labels: cnn
Automatic speech recognition
End-to-end Automatic Speech Recognition for Madarian and English in Tensorflow
Stars: ✭ 2,751 (+962.16%)
Mutual labels: cnn
DLSS
Deep Learning Super Sampling with Deep Convolutional Generative Adversarial Networks.
Stars: ✭ 88 (-66.02%)
Mutual labels: cnn
Caption generator
A modular library built on top of Keras and TensorFlow to generate a caption in natural language for any input image.
Stars: ✭ 243 (-6.18%)
Mutual labels: cnn
social-cnn-pytorch
Human Trajectory Prediction in Socially Interacting Crowds Using a CNN-based Architecture
Stars: ✭ 24 (-90.73%)
Mutual labels: cnn
Crnn Pytorch
Pytorch implementation of CRNN (CNN + RNN + CTCLoss) for all language OCR.
Stars: ✭ 248 (-4.25%)
Mutual labels: cnn
Selecsls Pytorch
Reference ImageNet implementation of SelecSLS CNN architecture proposed in the SIGGRAPH 2020 paper "XNect: Real-time Multi-Person 3D Motion Capture with a Single RGB Camera". The repository also includes code for pruning the model based on implicit sparsity emerging from adaptive gradient descent methods, as detailed in the CVPR 2019 paper "On implicit filter level sparsity in Convolutional Neural Networks".
Stars: ✭ 251 (-3.09%)
Mutual labels: cnn
WSCNNTDSaliency
[BMVC17] Weakly Supervised Saliency Detection with A Category-Driven Map Generator
Stars: ✭ 19 (-92.66%)
Mutual labels: cnn
Cnnforandroid
The Convolutional Neural Network(CNN) for Android
Stars: ✭ 245 (-5.41%)
Mutual labels: cnn
Lightnet
Efficient, transparent deep learning in hundreds of lines of code.
Stars: ✭ 243 (-6.18%)
Mutual labels: cnn
stylenet
A pytorch implemention of "StyleNet: Generating Attractive Visual Captions with Styles"
Stars: ✭ 58 (-77.61%)
Mutual labels: cnn
Facedetection
C++ project to implement MTCNN, a perfect face detect algorithm, on different DL frameworks. The most popular frameworks: caffe/mxnet/tensorflow, are all suppported now
Stars: ✭ 255 (-1.54%)
Mutual labels: cnn
deep-learning-coursera-complete
Deep Learning Specialization by Andrew Ng on Coursera - My Completed Coursework Repo - All 5 Courses
Stars: ✭ 104 (-59.85%)
Mutual labels: cnn
CNN Visualizer: Toolkit for Visualizing Units in Deep Convolutional Neural Networks
Introduction
This repository contains the codes and results for visualizing the units (or named as neuron or convolutional filters) inside deep CNN. Some of the codes have been used for the ICLR'15 paper Object Detectors Emerge in Deep Scene CNNs. You can use this toolkit with the naive caffe, with matcaffe and pycaffe compiled, or pyTorch. This toolkit includes the following functions:
(NEW!) PyTorch script:
-
pytorch_extract_feature.py: code to extract the CNN features at the selected layers of a CNN model for any given images. -
pytorch_generate_unitsegments.py: code to generate the visualization of all the units at the selected layer.
Matlab script:
-
extract_features.m: code to extract the CNN activation at the selected layers for any given images. -
generate_unitsegments.m: code to generate the visualization for all the units at the selected layer. -
unit_annotation: the results of unit annotation for Places-AlexNet and ImageNet-AlexNet. -
unit_segmentation: code to generate image segmentation for a single image using the synthetic receptive field. -
minimal_image: possion removal of the image content, to generate the minimal images.
PyTorch Environment
pip install -r requirement.txt
Download
- Clone the code from github
git clone https://github.com/metalbubble/cnnvisualizer.git
cd unitvisseg
- Download the image samples and the pretrained models
sh download_images.sh
sh download_pretrain.sh
Run
Properly set up the caffe path in the shell scripts, matlab scripts, and python scripts. We show the procedure to probe the unit semantics for alexnet trained on Places205 as follows:
- Generate the visualization for all the units in one layer of a given CNN. It gets the top ranked images for each unit then segment them using the approximate receptive field of the units. The results are output to result_segments. After it finishes, you could click
result_segments/places205_alexnet.htmlto see the visualization of unit segmentation.
matlab -nodisplay -r generate_unitsegments
Samples of unit segmentation are shown below. Each row indicates the images segmented by the receptive field of one unit.
- Generate the minimal image.
cd minimal_image
matlab -nodisplay -r demo_bases
Sample of the bases and images generated from combining the bases:
- Generate image segmentation using the synthetic receptive field of one unit.
cd unit_segmentation
matlab -nodisplay -r demo_unitsegments
It generates the image segmentation like this:
- Extract the CNN features from any layers of any CNNs for any images. It will output the CNN features for the selected layers.
matlab -nodisplay -r extract_features
References
If you find the toolkit useful, please cite our ICLR'15 paper:
Bolei Zhou, Aditya Khosla, Agata Lapedriza, Aude Oliva, and Antonio Torralba
Object Detectors Emerge in Deep Scene CNNs.
International Conference on Learning Representations (ICLR), 2015.
Note that the project description data, including the texts, logos, images, and/or trademarks,
for each open source project belongs to its rightful owner.
If you wish to add or remove any projects, please contact us at [email protected].