Custom shape predictor model trained to find 81 facial feature landmarks given any image

License: BSD 3-Clause "New" or "Revised" License

Python 100.00%

shape_predictor_81_face_landmarks's Introduction

81 Facial Landmarks Shape Predictor

This is a custom shape predictor model trained to find 81 facial feature landmarks given any image.

It's trained similar to dlib's 68 facial landmark shape predictor. In addition to the original 68 facial landmarks, I added an additional 13 landmarks to cover the forehead area. This allows for precise head detection and for image operations that require points along the top of the head, for example when placing a hat on someone's head.

The additional 13 points were extract using my fork of eos by patrikhuber: https://github.com/codeniko/eos. I continued to use the Surrey Face Model and made note of 13 points that I thought were the perfect fit for my use case. I made the modifications here, then ran it on the entire ibug large database of images to overwrite each image's 68 landmark coordinates with my 81 landmark coordinates. From here, the training for the shape predictor model can proceed using http://dlib.net/train_shape_predictor.py.html

Check out this live example video of 81 landmark detection in action: https://www.youtube.com/watch?v=mDJrASIB1T0

81 landmarks reference

shape_predictor_81_face_landmarks's People

Stargazers

Watchers

Forkers

shadowkun miwaliu qianji13 killerwuhan antonlinderer irentang lukaliu daojishigailvlun neo8245 pubfork guihui hollisjoe lonelygo kamingli1st hxhh panhiuchuen zilipeng gavin666github yaoxingtian zhengjianwei hmzjwhmzjw jinfei3459 yyyangup jackistom fastzhang lsmilesmile apexpredator1 paseam gengcauwong jiev daemonhaha eudora-jia hanshan123 kulasama elttaes hyb1234hi cicerolneto beknown-j yueyihua derek520 deftruth watsonwuh htmljsp jqueryjq ivanwilli98 moizkachwala happyllin wyuzyf courses-on-ai lmhlll nyaang tgsen edward-deltax duyhungtnn idomatheveryday sonfire186 lusc1996 luwei6896 19317362 sipwes superrichiesui richiesui vivianliangb slkwyy helloahuzw zjuming super7xia zxy2020 66pig autogyro saber190919 buptjz kaijung w303972870 lbb1993 adteven xrosliang delebash smane2000 skyrookieyu wuyx dreamer-shan xhair hakanaku1234 matt45m vincentsw1026 jiayu-ch15 vselfdom saulosilva chaerlo tian0zhi neilgautam holttechnologycorporation sohasamy easy-shu magicjva yueyedeai hellooword zhang11wu4 sinianyutian

shape_predictor_81_face_landmarks's Issues

Hi

hi, can u give me the 3D model index of the 3DMM

眼睛检测不准

81个点（20M）

68个点(96M)

Mesh

I have a question regarding full_mesh_example.png, how do you produce this?

Training Params

Hi Niko,

Could I trouble you for the training params you used to train your dlib model.

I'm trying to train a model on Helen at the moment and then will try on my custom forehead points (iBUG) further down the line.

I feel like I picked reasonable params according to documentation and articles out there but my train/test errors were well off. Would be very helpful for me to sense check.

Iain

An error appeared when running the demo

RuntimeError: Error deserializing object of type int

Model size

could you tell me about the size of the dataset you have trained this model on?
Davis kings dlib model was around 100mb , however yours is just 20mb
does the size depend on the data it trained on?

creating a new predictor

Hi, I want to create a predictor that just outputs 12 landmarks corresponding to eyes instead of all 68.How should my training xml file reflect to serve my purpose?
can i achieve it by only having 12 co-ordinates in the xml file?
Thanks in advance

Order of point is incorrect

Order of additional 13 points is incorrect

from 69 to 81

13 sfm indices

hi,thank u for sharing your result~
i want to find some feature points in my own project as you did,
how can i get the indices of the 13 sfm points?

how do i get addition 13 landmarks

training datasets only 68 landmarks, How to get the other 13 key points？Can you explain it more specificly?

Whats your license like?

How do I mention you if I fork this?

How to generate full mesh?

I see a full mesh figure. But how to generate the full mesh?

can't detect small face !

81 face landmarks with Caffe DNN

Hi sorry if this is abrupt.

First of all, thank you so much for the excellent work! Your trained data set really helped me so much!

But I'm wondering if it's possible that you could maybe share your source data so I can try to train it with Caffe DNN since it doesn't work well with Caffe DNN face detection.

I'll definitely share the trained model if I could train it successfully.
Thank you.

Question: Surrey Face Model Points

Hi @codeniko

Thanks for the .dat file super useful!

Sorry for my stupidity here, but this should be easy and taring my hair out here. Is there any documentation anywhere on the surrey face model and the numeration of the points within eos.

I've done all the hardwork and built a model in eos and projected the points into 2d. I can now choose some points and train my dlib model.

I've tried to just look and see the at output but there is no obvious pattern in how they are labelled.

Iain

How to make training prediction on one's own data set

Hello, I have a face data with 194 face key points. I want to use your network to train my data so that I can predict 194 face key points, but there is no explanation in your explanation. Please tell me how should I start?

codeniko / shape_predictor_81_face_landmarks Goto Github PK