https://github.com/cg1507/convnet-vis
ConvNet-Vis helps to visualize the Deep Convolutional Neural Networks with following methods.
https://github.com/cg1507/convnet-vis
convnet-zoo convolutional-neural-networks deconvolution deep-dream keras tensorboard tensorflow visualization-tools
Last synced: 3 months ago
JSON representation
ConvNet-Vis helps to visualize the Deep Convolutional Neural Networks with following methods.
- Host: GitHub
- URL: https://github.com/cg1507/convnet-vis
- Owner: CG1507
- License: mit
- Created: 2018-12-15T18:05:20.000Z (over 6 years ago)
- Default Branch: master
- Last Pushed: 2018-12-27T06:45:27.000Z (over 6 years ago)
- Last Synced: 2023-09-25T00:30:17.486Z (over 1 year ago)
- Topics: convnet-zoo, convolutional-neural-networks, deconvolution, deep-dream, keras, tensorboard, tensorflow, visualization-tools
- Language: Python
- Homepage:
- Size: 178 KB
- Stars: 3
- Watchers: 2
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# :milky_way: ConvNet-Vis
ConvNet-Vis helps to visualize the Deep Convolutional Neural Networks with following methods.
- Activation of image from each layer
- Deconvolution
- Deep-Dream
## Requirements:
* Tensorflow
* Keras
* numpy
* scipy
* h5py
* wget
* six
* scikit-image
## Run:
Very first time it will download the weights of the model you pick, so it requires an internet connection. Also you can pass custom model.
```python
import vis
# For pretrained model visualization
convnet_vis = vis.ConvNet_Vis(image_path="cat.jpg")
# For custom model visualization
# model= OR Keras Model obect
convnet_vis = vis.ConvNet_Vis(image_path="cat.jpg", model="final_model.hdf5")
```
> **NOTE:** Results will be stored in **vis** directory.
#### **Google-Colab** support added.
## For Tensorboard:
TensorBoard gives you flexibility to visualize all the test image on same model with brightness and contrast adustment.
```
tensorboard --logdir=
```
## Todo:
- [ ] Deep-dream support for all the model. **(Current Support only for InceptionV3)**
## License
[MIT](https://choosealicense.com/licenses/mit/)
## Author
[
](https://github.com/CG1507)
## Acknowledgement
:green_heart: [tfcnn_vis](https://github.com/InFoCusp/tf_cnnvis)