What a Deep Neural Network Thinks About Your Selfie

Stanford PhD student Andrej Karpathy trained a model overnight on a Tesla K40 to tell you how to take a better selfie photo. Convolutional Neural Networks are great at recognizing things, places and people in your personal photos, crops, traffic, various anomalies in medical images and all kinds of useful things.

But once in a while these powerful visual recognition models can also be warped for distraction, fun and amusement. Karpathy used a powerful, 140-million-parameter state-of-the-art Convolutional Neural Network, fed it 2 million selfies from the internet, and trained it to classify good selfies from bad ones.

For training, he used the Caffe deep learning framework and trained the model overnight on a Tesla K40 GPU.

Top 100 selfies
Top 100 selfies according to the ConvNet – recognize the trend?

He also created a Twitter bot (@deepselfie) that enables you to tweet your selfie and then it’ll auto rate it.

Read his entire blog to find out what makes a good selfie >>

About Brad Nemire

Brad Nemire
Brad Nemire is on the Developer Marketing team and loves reading about all of the fascinating research being done by developers using NVIDIA GPUs. Reach out to Brad on Twitter @BradNemire and let him know how you’re using GPUs to accelerate your research. Brad graduated from San Diego State University and currently resides in San Jose, CA. Follow @BradNemire on Twitter