yolov7 icon indicating copy to clipboard operation
yolov7 copied to clipboard

Yolov7 Architecture in Research Paper

Open thilinawee opened this issue 1 year ago • 3 comments

Hi, When I went through the research paper of yolov7, I came across this diagram which describes the architectural differences between some other networks and yolov7 itself. image

I wonder if anyone can explain to me the meanings of these notations. For example, what does 3x3, 2c, 2c, 2 stand for? I understand that 3x3 means a kernel_size = 3,3. However, I am not sure about the rest. Thanks ~Thilina

thilinawee avatar Sep 10 '22 16:09 thilinawee

kernel size, input channel, output channel(, group).

WongKinYiu avatar Sep 11 '22 04:09 WongKinYiu

Thanks, @WongKinYiu for the quick reply. Does group mean the number of groups in the convolution(group convolutions) or the number of such sequential blocks?

thilinawee avatar Sep 12 '22 06:09 thilinawee

They are equivalent, you could take a look ResNeXt paper. image

WongKinYiu avatar Sep 13 '22 10:09 WongKinYiu

I sketched the yolov7 architecture according to what I understood. Could you please point out where is E-ELAN block with group convolutions? Are the group convolutions used only for training? Thanks in advance.

Overall Architecture

image

ELAN

image

CSPSPP

image

B3

image

B2

image

thilinawee avatar Sep 21 '22 17:09 thilinawee

Have you received any feedback on your draft?

RomanczuG avatar Oct 16 '22 21:10 RomanczuG

@RomanczuG No, I haven't :(

thilinawee avatar Oct 21 '22 06:10 thilinawee

You could take a look page 16 of the paper.

WongKinYiu avatar Oct 21 '22 07:10 WongKinYiu

So you implemented E-Elan just in yolov7-E6E?

PascAlex avatar Oct 21 '22 08:10 PascAlex

@thilinawee Hello, i'm new to yolo algo. Can you explain more about RepConv layer?

Firdaus909 avatar Feb 25 '23 03:02 Firdaus909