People and face detection in ImageNet

[Henrik L]

Hi guys,

I'm just starting out with a research project that I am hoping Caffe would be well suited for. I want to gather stats on what kind of images the major news outlets use on their frontpages.

I have installed caffe and fed it with the pretrained imagenet model, and I have it up and running. I even modified the web app so that you can point it at an URL and specify the class name of the images on that URL, and it will download them and classify them.

BUT: I am having some trouble with the categories that are returned back from the pretrained imagenet model. I'm not getting any categories like "faces", "people", "outdoors" and other general categories that would be very beneficial in a statistical summary.

So, what I'm wondering about is: Does the ImageNet model have these categories? If not, are there any other datasets that are readily available that I can push into Caffe, that are better suited for my needs?

Look forward to hearing from you!