search

anatolix / keras_Realtime_Multi-Person_Pose_Estimation

Keras version of Realtime Multi-Person Pose Estimation project
Other
118 stars 47 forks source link

How can I use the same implementation for a single person 2D pose estimation? #27

Closed abinjoabraham closed 6 years ago

abinjoabraham commented 6 years ago

@anatolix First of all thanks a lot for your work. When I compare with the michalfaber code I find this one more simple. Being a new guy to this field of artificial intelligence this implementation will be a lot helpful for me for the future references. Can I modify the same implementation to detect the pose for a single person rather than the multiperson?

murrayLuke commented 6 years ago

This should work with one person as well as multiple people.

abinjoabraham commented 6 years ago

@murrayLuke Thanks for your input. I understand that this will work for single person as well. But was just thinking whether I can reduce the training set data so that it need to work only for the single person. By this way I can easily create a smaller version which I can train it in my mac book pro 2017 edition with no nVidea GPU. :)

anatolix commented 6 years ago

Imho no chance you could train it without GPU.

abinjoabraham commented 6 years ago

Sorry for this lame question. Finally after going through some papers and implementations I finally think that the same can be done with only the heat maps. And the response time also will be better for single person as the weights will be less for the same as we reduce the complexity of the system.

abinjoabraham commented 6 years ago

Hence I am closing this question. Thanks for your valuable time in replying to this question.