This project is maintained by jhoffman


Welcome. LSDA is a framework for large scale detection through adaptation. We combine adaptation techniques with deep convolutional models to create a fast and effective large scale detection network.

7.5K Detector Model

We're releasing a 7604 category detector model for use within Caffe. The categories correspond to the 7404 leaf nodes from the ImageNet dataset, as well as 200 stonger detectors that are available with bounding box data from the ILSVRC2013 challenge dataset.

Download the pre-trained model. Then run the demo script in the folder.

We also have a fork of Caffe with the ability to run a 7.5K detector in less than a second per image. Check out the fast lsda project page!

Citing LSDA

   Author = {Judy Hoffman and Sergio Guadarrama and Eric Tzeng and 
   		Ronghang Hu and Jeff Donahue and Ross Girshick and 
	   	Trevor Darrell and Kate Saenko},
   Title = { {LSDA}: Large Scale Detection through Adaptation},
   Year  = {2014},
   booktitle = {Neural Information Processing Systems (NIPS)}

Authors and Contributors

Judy Hoffman
Sergio Guadarrama

Eric Tzeng

Jeff Donahue

Ross Girshick

Trevor Darrell

Kate Saenko

This page and the corresponding software is maintained by Judy Hoffman (@jhoffman) and Sergio Guadarrama (@sguada). The work is supported by the Berkeley vision group and BVLC.