search

yeephycho / tensorflow-face-detection

A mobilenet SSD based face detector, powered by tensorflow object detection api, trained by WIDERFACE dataset.
Apache License 2.0
768 stars 264 forks source link

Config needed #3

Closed vaishnavm217 closed 6 years ago

vaishnavm217 commented 6 years ago

Hi, I am training the same model on WIDERface with the coco default config. but running eval.py is giving warning for

no ground truth in the following classes [1,2]

i am using the label map provided by your repo. even went through some of the issues of this problem, still i am getting the same error. Can you post config you used so i can decode this problem?

yeephycho commented 6 years ago

I am sorry but I'm currently lack of bandwidth to respond your request. But I think your request is somehow reasonable, my suggestion for you is to go through the tutorial from google object detection api in tensorflow. I believe that after you finished all the related materials, you are to able to train your customed model. It would take to much time.

Good luck~

vaishnavm217 commented 6 years ago

Sorry to intrude, I just want to ask did you change the whole config file, like parameters and minimum size or just changed the PATH_TO_CONFIGURE ?

yeephycho commented 6 years ago

I think I changed necessary part of the config file.

taewookim commented 6 years ago

@yeephycho

In your face_label_map, i noticed background as one of the labels. Why was that included? I don't recall seeing that in the TF object detection api tutorial (i.e. the custom pet detector label map)

yeephycho commented 6 years ago

You can ignore the label number or change it to any number you like.