-
Notifications
You must be signed in to change notification settings - Fork 21.3k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Why should this line "assert(l.outputs == params.inputs) " in line 281 of parser.c #236
Comments
Hi there! Did you by any chance solve the problem? @MaeThird |
@MaeThird @jktzes According to my personal experience |
what is "num"??? |
@dee6600 It's set in your configuration file. For yolo3 the default the 9 but you can change this depending on your needs. |
I am working with yolov2, For three categories. so, filters = (3+5)*5 seems to work. |
@dee6600 num is the amount of anchor-pairs, so if you increase the amount of anchors you should increase num as well. |
@TheMikeyR thank you for insight. But i am kind of noob here hehe. I dont know what is anchor-pairs. |
Hello there, |
I thought num was the number of classes that can overlap in one window. After reading #568 I think its a new concept replacing simple grid used in yolo v1. |
in the yolov3-voc.cfg file there are three [yolo] block. you have to change the filter size for each of the [convolutional] block filter size before all the three [yolo] block as follow - [convolutional] [yolo] |
In your .cfg make sure the "classes = num of your trained object" in each layers |
Hi. I am using Yolov3. |
If your try to run the training from terminal, after you've changed the filters and classes, try to exit the terminal and start it again. It may work like that. And do not forget to save the changes that you made in .cfg file |
What's the job of this assert
assert(l.outputs == params.inputs)
in line 281 ofparser.c
?In my opinion the
parse_region
is doing the assembling works of the regional proposal layer,in which the output is ordinarily asnumber_of_proposes*(classes + coordinates +1)
,although inconvolutional_layer
the output is ordinarily calculated ash*w*filters
.I mean there is no logical equal between input size and output size of regional proposal layer.
Can any explanation ?
The text was updated successfully, but these errors were encountered: