YOLOv4-tiny size mismatch error

ajaykumaar · September 13, 2020, 10:31am

Hii,
I’m trying to create an object detector with YOLOv4-tiny and I got the weights and cfg file from the below links.

https://github.com/AlexeyAB/darknet/releases/download/darknet_yolo_v4_pre/yolov4-

AlexeyAB/darknet/blob/master/cfg/yolov4-tiny.cfg

[net]
# Testing
#batch=1
#subdivisions=1
# Training
batch=64
subdivisions=1
width=416
height=416
channels=3
momentum=0.9
decay=0.0005
angle=0
saturation = 1.5
exposure = 1.5
hue=.1

learning_rate=0.00261
burn_in=1000

This file has been truncated. show original

I get an error while loading the weights to Darknet model.

How do i resolve this?

Thanks in advance

ptrblck · September 15, 2020, 10:26am

Unfortunately I don’t know how to resolve this issue as I’m not familiar with your code, but the error is most likely raised by the view_as operation, which tries to view the parameter in an invalid shape.
You should thus either fix the slicing operation on weights or the conv_layer.weight shape.

ajaykumaar · September 21, 2020, 6:58am

Hi @ptrblck,
Thank you for ur reply. Yeah right, the issue was due to invalid shape because of a bug in the architecture. Now it works fine.
Thank you.

Sathish_Basavaraju · June 23, 2021, 8:19pm

Hi @ajaykumaar ,

I am facing the same issue. Could you please explain, how you solved the problem?
Thank you.