https://github.com/geyang/image-segmentation

Getting familiar with image segmentation
https://github.com/geyang/image-segmentation

deep-learning playground segmentation tensorflow

Last synced: 3 months ago
JSON representation

Getting familiar with image segmentation

• Host: GitHub
• URL: https://github.com/geyang/image-segmentation
• Owner: geyang
• Created: 2018-02-02T19:53:49.000Z (over 7 years ago)
• Default Branch: master
• Last Pushed: 2018-02-04T07:12:19.000Z (over 7 years ago)
• Last Synced: 2025-01-10T12:58:30.326Z (5 months ago)
• Topics: deep-learning, playground, segmentation, tensorflow
• Language: Python
• Homepage:
• Size: 12.9 MB
• Stars: 0
• Watchers: 1
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

        
---

tags: ['deep-learning', 'playground']

---

# Playground/image-segmentation

This `TensorFlow` implementation of ENet is forked from @fregu856. You can find his 

original code here: [@fregu856/segmentation.git](https://github.com/fregu856/segmentation), which is implemented based on 

the pyTorch version at https://github.com/e-lab/ENet-training

## Show me the money!

Here is a gif of the trained network: click on [the gif](https://www.youtube.com/watch?v=4OB9a3KgCXM) :point_left: to go to the youtube video.

[![ENet segmentation cityscapes stuggart 00](./figures/cityscapes_stuggart_00_short.gif)](https://www.youtube.com/watch?v=4OB9a3KgCXM)

## TODOs

- parallelize training to use more GPUs

### ENet Paper: 

The original paper for this model could be found here: 

[2016 Adam Paszke, et.al., ENet: A Deep Neural Network Architecture for Real-Time Semantic Segmentation

](https://arxiv.org/pdf/1606.02147.pdf)

## Common Problems

If you encounter the "No gradient defined for operation 'MaxPoolWithArgmax_1' (op type: MaxPoolWithArgmax)", you just need to use a GPU tensorflow instance, b/c the max_pool_with_argmax operator is not defined for the CPU.

                                             k