- Result: Visualization Page, arxiv
- Support models from caffe (June 2016) and matconvnet (1.0-beta12, May 2015)
(caffe support: AlexNet/VGG-16/NIN/GoogleNet; matconvet support: AlexNet)
- link and put under models/
- Edit param_init.m: caffe/matcaffe location
-
Visualize single neurons from CNN model:
-
Edit and run: V_neuronInv.m
-
Image Completion
-
Download object topics link and put under data/fc7-topic/
-
Edit and run: V_inpaint.m
-
Feature Inversion, same as deep-goggle
-
Edit and run: V_featInv.m
- Visualize Intra-class variation
- Visualize hierarchical binary CNN code
- 2016.11: bug fixed
- 2016.08: code organization
- update caffe version to June 2016
- add links to compatible CNN models
- 2016.04: add object insertion
- object insertion (V_app_inpaint.m)
- 2015.07: initial release
- support caffe (July 2015) and matconvnet (1.0-beta12, May 2015)
- single neuron visualization (V_neuron_single.m)
The code is heavily based on aravindhm's deep-goggle