How to use deep scene features for new scene classification training?

CSAILVision / places365

The Places365-CNNs for Scene Classification

http://places2.csail.mit.edu/

MIT License

1.9k stars 534 forks source link

How to use deep scene features for new scene classification training? #31

Closed ghost closed 6 years ago

ghost commented 6 years ago

If we want to test the performance of the deep features of Places365-CNNs, how to set up new training for new dataset such as Scene15?

ghost commented 6 years ago

I rarely use Caffe, please correct me if I am wrong. For a fine-tune task of smaller dataset, in .prototxt file of pretrained model, we change the names of top layers, and set to zero the "lr_mult" of other layers. In train_placesCNN.py, what do we pass to the --pretrained argument?

zhoubolei commented 6 years ago

google the tutorial for fine-tuning caffemodel.