https://github.com/christiankaiser/segnet.tf

SegNet with Tensorflow
https://github.com/christiankaiser/segnet.tf

Last synced: 3 months ago
JSON representation

SegNet with Tensorflow

• Host: GitHub
• URL: https://github.com/christiankaiser/segnet.tf
• Owner: christiankaiser
• Created: 2018-02-26T06:41:58.000Z (over 6 years ago)
• Default Branch: master
• Last Pushed: 2018-02-26T06:43:37.000Z (over 6 years ago)
• Last Synced: 2024-05-19T23:36:14.449Z (6 months ago)
• Language: Python
• Size: 12.7 KB
• Stars: 1
• Watchers: 2
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

        
# SegNet with Tensorflow

SegNet implementation for Tensorflow with instructions to run the model on Floydhub. This implementation is based on [github.com/tkuanlun350/Tensorflow-SegNet](https://github.com/tkuanlun350/Tensorflow-SegNet).

The implementation is a slightly modified [SegNet](http://arxiv.org/abs/1511.00561) in Tensorflow trained on the CamVid dataset (see section Dataset for the details).

In this implementation the original upsampling method is replaced simply by a deconv (or conv-transpose) layer (without pooling indices), because indice unravelling is not yet unavailable in Tensorflow (see corresponding [issue](https://github.com/tensorflow/tensorflow/issues/2169)).

## Dataset

We use here the Alex Kendall's CamVid dataset originally available from the [SegNet Tutorial](https://github.com/alexgkendall/SegNet-Tutorial) and available on Floydhub ffrom [humpelandi/datasets/camvid/1](https://www.floydhub.com/humpelandi/datasets/camvid).

## Run it on Floydhub

Create the project on Floydhub:

```

floyd init segnet

```

Training the network:

```

floyd run --gpu --env tensorflow-1.3 --data humpelandi/datasets/camvid/1:/CamVid 'python train.py'

```

It is possible to finetune an existing model:

```

# Not yet implemented

```

Testing is done with:

```

floyd run \

      --gpu \

      --env tensorflow-1.3 \

      --data humpelandi/datasets/camvid/1:/CamVid \

      --data humpelandi/projects/segnet/1/output:/model \

      'python test.py'

```