Permalink
Branch: master
Find file Copy path
Fetching contributors…
Cannot retrieve contributors at this time
58 lines (33 sloc) 3.2 KB

SegNet Model Zoo

This page lists a number of example SegNet models in the SegNet Model Zoo. NOTE: all Bayesian SegNet models can be tested as SegNet models (for example by using the webcam demo) by removing the line sample_weights_test: true on all Dropout layers, and setting batch size of 1.

Driving Web Demo

This example model is used in the SegNet webdemo [http://mi.eng.cam.ac.uk/projects/segnet/]. It is trained to classify road scenes into 12 classes.

Model file: segnet_model_driving_webdemo.prototxt

Weights can be downloaded from this link: [http://mi.eng.cam.ac.uk/~agk34/resources/SegNet/segnet_weights_driving_webdemo.caffemodel]

CamVid

These models have been trained for road scene understanding using the CamVid dataset.

SUN

These models have been trained for indoor scene understanding using the SUN RGB-D dataset.

The model definition file used for training can be found here train_segnet_sun.prototxt

We have also trained a model for a 224x224 pixel input:

Pascal VOC

These models have been trained on the Pascal VOC 2012 dataset .

This model is based on the Dropout enc-dec variant and is designed for a 224x224 pixel input.

CityScapes

This model finetuned the webdemo weights using the CityScapes dataset (11 class version).

License

These models are released under a creative commons license which allows for personal and research use only. For a commercial license please contact the authors. You can view a license summary here: http://creativecommons.org/licenses/by-nc/4.0/

Contact

Alex Kendall agk34@cam.ac.uk Cambridge University