search

zzeitt / forGestureRecognition

Recognize gestures using a simple webcam.
30 stars 11 forks source link

question about tensorflow version #2

Open RKLho opened 4 years ago

RKLho commented 4 years ago

I would like to use your code for gesture detection. But I have some problem with load_inference_graph function. I guess it is a problem of tensorflow version. Could you share your training tensorflow version for frozen_inference_graph_2.pb and frozen_inference_graph.pb?

zzeitt commented 4 years ago

Hi @RKLho , this project is a redevelopment based on @victoria's handtracking, I didn't train the model myself, you could check the original repo for trainning detail.

As for the tensorflow version, I am using tensorflow==1.13.1.

yangqu2000 commented 4 years ago

Hi @zzeitt , when I run your detect_image.py, I found that there are only six parameters inside "recognizer_utils.drawBoxOfROI()" function but in recognizer_utils.py, the original should be seven parameters. It seems that the "margin_ratio" is missing. But when I add margin_ratio to the detect_image.py and ran it, error happened as "too many values to unpack (expected 2)". Could you please check this problem for me? Seem question also happens in detect_multi_threaded.py. Lots of thanks!

zzeitt commented 4 years ago

I replicated your problem and found that I missed one return value and one argument for function drawBoxOfROI() in detect_image.py. (BTW, detect_webcam.py is all good.)

I am sorry that the codes are not under maintaining currently because I could not spare time to do this. So there might be some bugs:bug: here and there...

Improvement for this project is welcome!