hello, can I use your trained model for age estimating ? for Chinese face..

[9crk]

I'v got money short and have no GPU computer by hand... and I dont want to cost days on trainning the model, and I dont have huge amount of pictures.

It seems not much precise in my test.. the input is the face I cuted from pictures. it both turn to 0 when I put in a 1year baby and a 70 year old lady.

net_def='../models/c9e99062283c719c03de/deploy_age.prototxt' net_mod='../models/c9e99062283c719c03de/age_net.caffemodel' net = caffe.Classifier(net_def,net_mod , image_dims=[256,256], mean=None, input_scale=None, raw_scale=255.0, channel_swap=[2,1,0]) inputs = [caffe.io.load_image(filename)] predictions = net.predict(inputs,0) print predictions.argmax()

[9crk]

or if you have any test imageset ? pls give it to me.

[9crk]

after loading the mean file,I got a right classification. you might need to declear this https://github.com/BVLC/caffe/wiki/Model-Zoo#models-for-age-and-gender-classification

mention that age estimate should load the mean file too.

btw, is the gender model load the same mean file too ?

[GilLevi]

Hi,

Yes, the gender model uses the same mean file.

Gil

[9crk]

thank you for reply. I got a bad result when using the caffemodel you give on googledrive. this is all my code:

net_def='../models/c9e99062283c719c03de/deploy_age.prototxt'
net_mod='../models/c9e99062283c719c03de/age_net.caffemodel'
net_mean='../models/c9e99062283c719c03de/mean.binaryproto'

proto_data = open(net_mean, "rb").read()
a = caffe.io.caffe_pb2.BlobProto.FromString(proto_data)
mean  = caffe.io.blobproto_to_array(a)[0]

net = caffe.Classifier(net_def,net_mod ,
            image_dims=[256,256], mean=mean,
            input_scale=None, raw_scale=255,
            channel_swap=[2,1,0])
def work(filename):
    inputs = [caffe.io.load_image(filename)]
    start = time.time()
    predictions = net.predict(inputs,0)
    print("Done in %.2f s." % (time.time() - start))
    print predictions
    return predictions.argmax()
if __name__ == '__main__':
    print work(sys.argv[1]);

once the man's face in the picture is smooth. then it turns to be the age belong to a child... I want to know if there is some certain condition when using this model. like the resolution

or is this caffemodel file is not well trained yet?

this image turns to be ![Uploading WX20170327-134034.png…]() 8-12year old.

[9crk]

![Uploading WX20170327-134034.png…]()

[9crk]

oh..I miss-clicked it...

[GilLevi]

Can you upload the image?

[9crk]

Its nice of you to reply..thanks!

I got trouble to upload pictures to GitHub in China.....

as you see. the two picture is the same face. but one turns a male and one to female...

[9crk]

and this is the face turned to be 8 years old..

[GilLevi]

When I ran it on my laptop I actually got 'Male" in both of those image. I also got that the bottom face is 8 years old. The model is not perfect, but given a larger training set, you can improve it.

[gittigxuy]

the accuracy seems to be not good.So,how can I improve the accuracy?Should I change the network structure? If I just increase the number of the train set picture.will the result be better?the last question is that if I increase the nunber of the train set picture,can I continue to use caffemodel you provide to me.how can I improve the accuracy rate of the caffemodel?