search

szaza / android-yolo-v2

Android YOLO real time object detection sample application with Tensorflow mobile.
Do What The F*ck You Want To Public License
227 stars 86 forks source link

LegacyFaceDetectMapper: onFaceDetection - Ignored some incoming faces sinceface detection was disabled #14

Open aaronhan92 opened 5 years ago

aaronhan92 commented 5 years ago

Hi szaza, do you have any idea how to enable the onFaceDetection so that my face objection can work

szaza commented 5 years ago

What do you mean under onFaceDetection? This project does not contain any face detection feature. Did you extended it? Can you provide me more details about your work?

aaronhan92 commented 5 years ago

Goal: To implement face detection and facial expression recognition using Darkflow. Please kindly comment whether if my implementation flow works as follow:

1) Git clone the darkflow.git

2) Download pre-trained tiny-yolo-widerface.cfg and tiny-yolo-widerface_final.weights from https://drive.google.com/file/d/1wAy6-XWDBHMMawgPfKv2hp_Q5NBYqK1n/view (Able to detect face with bounding box that trained on WiderFace dataset)

3) Prepare my own small dataset (also copy from Widerface) using LabelImg software tool to crop and label each facial expression such as happy and sad etc.

4) Start to train my model using this command 'python flow --model cfg/tiny-yolo-widerface.cfg --train --trainer adam --annotation train_set/Annotation --dataset training_set/Images'.

5) Generate the pb file around 180,000KB size using this command 'python flow --model cfg/tiny-yolo-widerface.cfg --load 8000 --savepb

6) Place the generated pb file to the android asset folder and run the app. Realize that face detection does not work due to error message 'LegacyFaceDetectMapper: onFaceDetection - Ignored some incoming faces sinceface detection was disabled' (I think that the problem lies on Camera API 2)

szaza commented 5 years ago

Does your generated model work with darkflow?

aaronhan92 commented 5 years ago

Yes, I'm able to use the generated model to detect my face via web cam. Moreover, after generated the pb file using this command 'python flow --model cfg/tiny-yolo-widerface.cfg --load 8000 --savepb', I placed it to the asset folder and start to run the android studio.

However, during debug phase, I encountered this error message 'LegacyFaceDetectMapper: onFaceDetection - Ignored some incoming faces sinceface detection was disabled' (I think the problem exist is due to Camera API 2)