https://github.com/mbojarski/VisualBackProp

VisualBackProp - visualization method for convolutional neural networks
https://github.com/mbojarski/VisualBackProp

Last synced: 3 months ago
JSON representation

VisualBackProp - visualization method for convolutional neural networks

• Host: GitHub
• URL: https://github.com/mbojarski/VisualBackProp
• Owner: mbojarski
• Created: 2016-11-17T02:47:54.000Z (over 7 years ago)
• Default Branch: master
• Last Pushed: 2016-11-19T02:48:35.000Z (over 7 years ago)
• Last Synced: 2024-01-16T23:13:04.106Z (5 months ago)
• Language: Lua
• Size: 1.24 MB
• Stars: 45
• Watchers: 5
• Forks: 13
• Open Issues: 3
• Metadata Files:
  • Readme: README.md

Lists

• awesome-stars - VisualBackProp - visualization method for convolutional neural networks | mbojarski | 45 | (Lua)

README

        
# VisualBackProp

VisualBackProp - visualization method for convolutional neural networks

# Description

Detailed description of the VisualBackProp can be found in:

https://arxiv.org/abs/1611.05418

Content:

* inputImages - folder with input images

* outputImages - folder for output images

* vis.lua - implementation of VisualBackProp 

* run.lua - loads images, runs visualization method, saves images

* model.t7b - model trained to predict steering wheel angle

# Usage

`mkdir outputImages`

`qlua run.lua`

Script generates set of 3 images for each input image. Each set contain:

* Input image with visualization mask overlaid in red

* Averaged feature maps

* Intermediate masks