What about the possibility for users to send the base64 encoded images in the request and get the response of Face detection?
Since base64 is a pretty common way of storing the images so I was just wondering if it will be wise for the api to accept it as an image in the request.18 votes
- Train user from facebook profile.
- Use all photo tags (remove tags that don't look like user)
- Face.com had this feature, it was 95% accurate.
- Train facebook friends faces.17 votes
Provide a mechanism to 1) remove a userID, and all associated trained pictures 2) remove all trained pictures for a userID14 votes
Say I have a database with the various kinds of the parts of face: eyebrows, noses, lips, cheekbones, etc.
All of them are categorized, for example: http://www.indianbeautyforever.com/2012/09/different-types-of-eyebrows-and-face.html
And I want to parse face and detect a type of each and any part of the face.7 votes
The gui should allow users to remove trained faces aswell as show all saved data and give an opportunity to batch modify it.6 votes
As title. Estimate the ethnicity of detected faces.6 votes
We plan to implement this feature but we need more info on which ethnicity groups you expect to get from our API. If you vote for this feature, please write a comment about that.
No faces detected on the following picture: http://25.media.tumblr.com/tumblr_lfxhv2mFTN1qc2r4co1_500.jpg
This is a bug, or lack of recognition quality.
No faces are detected on the following picture: http://25.media.tumblr.com/tumblr_lfxhv2mFTN1qc2r4co1_500.jpg
There are 6 faces in it. And 4 of them are with a little angle.6 votes
Well, you have not tried the recognition yet. This is only face detection part.
After thorough examination of the image provided by you we found that it contains: 1 top of the head, 4 half-profile faces, and 1 full-profile face.
3 of the 4 half-profile faces are too small, our algorithms can easily detect them but we intentionally avoid this because we cannot meaningfully use them during the recognition. Additionally, detection of such small faces adds computation heaviness which we plan to compensate for our commercial plan users.
1 top of the head will not be detected by face detection algorithms.
1 full-profile face is not detected for the same reason as the small faces. We are still discussing internally how to offer this functionality.
It'd be great if you had a publicly available status page so in the rare occasions when outages occur we can verify that it is not our end and that you are working on it.6 votes
We will think of how our documentation can be improved. Examples based documentation is one of the possibilities. Thank you for the feedback.
Algorithm available on face-shape.com3 votes
Unfortunately, our face landmarks detector does not detect top of the forehead, which is needed for a face shape recognition algorithm. Maybe you could reuse eyebrows or eyes for the same purpose. Check all 68 landmarks that are returned by our API for similar tasks.
It will be very useful if we can remove tags via uid not tid. Because uid is writable and developer has more control over it despite of tid. When we want to remove a tag we should call additional tags/get to get specified tid and then delete it and then call additional faces/train. I think this should be only one request with removing tag via uid and auto train.
This will be very useful for developers and SkyBiometry : )
Hope we can use it very soon.3 votes
This is pretty self explanatory. While it's easy to do this at the app level it creates a lot of overhead especially for things with relatively high frequency.2 votes
How about detecting the hair of the face.1 vote
what about real time face recognition from video input?1 vote
Hourly, Daily and Monthly limits.1 vote
- Don't see your idea?