Convert yolov2 model

I want to convert yolov2 file trained by myself
but I faced with problem.
I get a wrong detection picture

cfg, weight file trained myself is successful in window
is there anyone who has this problem?


  1. Which demo you use for yolov2 ?
  2. How many classification models is this ?


  1. i use demo ’ aml_npu_sdk_6.4.3 ', ’ DDK_6.4.3 ’
  2. i only one detect ’ person ’ and I did classification test to train data.

@111111 Which one pre- and post-processs you used ?

And if you just need to detect person , maybe yolo not a good chooise.

sry, i don’t understand about ’ pre, post-process ’ you talked.

if you mean ’ pre, post process ', [ linux_sdk ] - [ demo ] - [ inceptionv3 ] ?
maybe, it has a diffrerence? when i converted yolov2 model and yolov3 model

@111111 I’m not sure if a single category 1 is feasible, and the recognition results of yolov3 that some users have done before are also wrong. The yolo design is not designed to identify people individually, I suggest you try the model that identifies people individually, and it will be lighter