search

AlexeyAB / darknet

YOLOv4 / Scaled-YOLOv4 / YOLO - Neural Networks for Object Detection (Windows and Linux version of Darknet )
http://pjreddie.com/darknet/
Other
21.75k stars 7.96k forks source link

layer before convolutional layer must output image: no error #4978

Open ghost opened 4 years ago

ghost commented 4 years ago

image

Hello,i'm having this error,my network size is 608x608,how can i solve this? thank you

WongKinYiu commented 4 years ago

could you check the size of 11th layer. image the output of 93th layer is 0x0x0.

ghost commented 4 years ago

[net]

Testing

batch=64 subdivisions=64

Training

batch=64

subdivisions=16

width=608 height=608 channels=3 momentum=0.9 decay=0.0005 angle=0 saturation = 1.5 exposure = 1.5 hue=.1 flip=0 learning_rate=0.001 burn_in=1000 max_batches = 172000 policy=steps steps=137600,154800 scales=.1,.1

[convolutional] batch_normalize=1 filters=32 size=3 stride=1 pad=1 activation=leaky

Downsample

[convolutional] batch_normalize=1 filters=64 size=3 stride=2 pad=1 activation=leaky

[convolutional] batch_normalize=1 filters=32 size=1 stride=1 pad=1 activation=leaky

[convolutional] batch_normalize=1 filters=64 size=3 stride=1 pad=1 activation=leaky

[shortcut] from=-3 activation=linear

Downsample

[convolutional] batch_normalize=1 filters=128 size=3 stride=2 pad=1 activation=leaky

[convolutional] batch_normalize=1 filters=64 size=1 stride=1 pad=1 activation=leaky

[convolutional] batch_normalize=1 filters=128 size=3 stride=1 pad=1 activation=leaky

[shortcut] from=-3 activation=linear

[convolutional] batch_normalize=1 filters=64 size=1 stride=1 pad=1 activation=leaky

[convolutional] batch_normalize=1 filters=128 size=3 stride=1 pad=1 activation=leaky

[shortcut] from=-3 activation=linear

Downsample

[convolutional] batch_normalize=1 filters=256 size=3 stride=2 pad=1 activation=leaky

[convolutional] batch_normalize=1 filters=128 size=1 stride=1 pad=1 activation=leaky

[convolutional] batch_normalize=1 filters=256 size=3 stride=1 pad=1 activation=leaky

[shortcut] from=-3 activation=linear

[convolutional] batch_normalize=1 filters=128 size=1 stride=1 pad=1 activation=leaky

[convolutional] batch_normalize=1 filters=256 size=3 stride=1 pad=1 activation=leaky

[shortcut] from=-3 activation=linear

[convolutional] batch_normalize=1 filters=128 size=1 stride=1 pad=1 activation=leaky

[convolutional] batch_normalize=1 filters=256 size=3 stride=1 pad=1 activation=leaky

[shortcut] from=-3 activation=linear

[convolutional] batch_normalize=1 filters=128 size=1 stride=1 pad=1 activation=leaky

[convolutional] batch_normalize=1 filters=256 size=3 stride=1 pad=1 activation=leaky

[shortcut] from=-3 activation=linear

[convolutional] batch_normalize=1 filters=128 size=1 stride=1 pad=1 activation=leaky

[convolutional] batch_normalize=1 filters=256 size=3 stride=1 pad=1 activation=leaky

[shortcut] from=-3 activation=linear

[convolutional] batch_normalize=1 filters=128 size=1 stride=1 pad=1 activation=leaky

[convolutional] batch_normalize=1 filters=256 size=3 stride=1 pad=1 activation=leaky

[shortcut] from=-3 activation=linear

[convolutional] batch_normalize=1 filters=128 size=1 stride=1 pad=1 activation=leaky

[convolutional] batch_normalize=1 filters=256 size=3 stride=1 pad=1 activation=leaky

[shortcut] from=-3 activation=linear

[convolutional] batch_normalize=1 filters=128 size=1 stride=1 pad=1 activation=leaky

[convolutional] batch_normalize=1 filters=256 size=3 stride=1 pad=1 activation=leaky

[shortcut] from=-3 activation=linear

Downsample

[convolutional] batch_normalize=1 filters=512 size=3 stride=2 pad=1 activation=leaky

[convolutional] batch_normalize=1 filters=256 size=1 stride=1 pad=1 activation=leaky

[convolutional] batch_normalize=1 filters=512 size=3 stride=1 pad=1 activation=leaky

[shortcut] from=-3 activation=linear

[convolutional] batch_normalize=1 filters=256 size=1 stride=1 pad=1 activation=leaky

[convolutional] batch_normalize=1 filters=512 size=3 stride=1 pad=1 activation=leaky

[shortcut] from=-3 activation=linear

[convolutional] batch_normalize=1 filters=256 size=1 stride=1 pad=1 activation=leaky

[convolutional] batch_normalize=1 filters=512 size=3 stride=1 pad=1 activation=leaky

[shortcut] from=-3 activation=linear

[convolutional] batch_normalize=1 filters=256 size=1 stride=1 pad=1 activation=leaky

[convolutional] batch_normalize=1 filters=512 size=3 stride=1 pad=1 activation=leaky

[shortcut] from=-3 activation=linear

[convolutional] batch_normalize=1 filters=256 size=1 stride=1 pad=1 activation=leaky

[convolutional] batch_normalize=1 filters=512 size=3 stride=1 pad=1 activation=leaky

[shortcut] from=-3 activation=linear

[convolutional] batch_normalize=1 filters=256 size=1 stride=1 pad=1 activation=leaky

[convolutional] batch_normalize=1 filters=512 size=3 stride=1 pad=1 activation=leaky

[shortcut] from=-3 activation=linear

[convolutional] batch_normalize=1 filters=256 size=1 stride=1 pad=1 activation=leaky

[convolutional] batch_normalize=1 filters=512 size=3 stride=1 pad=1 activation=leaky

[shortcut] from=-3 activation=linear

[convolutional] batch_normalize=1 filters=256 size=1 stride=1 pad=1 activation=leaky

[convolutional] batch_normalize=1 filters=512 size=3 stride=1 pad=1 activation=leaky

[shortcut] from=-3 activation=linear

Downsample

[convolutional] batch_normalize=1 filters=1024 size=3 stride=2 pad=1 activation=leaky

[convolutional] batch_normalize=1 filters=512 size=1 stride=1 pad=1 activation=leaky

[convolutional] batch_normalize=1 filters=1024 size=3 stride=1 pad=1 activation=leaky

[shortcut] from=-3 activation=linear

[convolutional] batch_normalize=1 filters=512 size=1 stride=1 pad=1 activation=leaky

[convolutional] batch_normalize=1 filters=1024 size=3 stride=1 pad=1 activation=leaky

[shortcut] from=-3 activation=linear

[convolutional] batch_normalize=1 filters=512 size=1 stride=1 pad=1 activation=leaky

[convolutional] batch_normalize=1 filters=1024 size=3 stride=1 pad=1 activation=leaky

[shortcut] from=-3 activation=linear

[convolutional] batch_normalize=1 filters=512 size=1 stride=1 pad=1 activation=leaky

[convolutional] batch_normalize=1 filters=1024 size=3 stride=1 pad=1 activation=leaky

[shortcut] from=-3 activation=linear

######################

[convolutional] batch_normalize=1 filters=512 size=1 stride=1 pad=1 activation=leaky

[convolutional] batch_normalize=1 size=3 stride=1 pad=1 filters=1024 activation=leaky

[convolutional] batch_normalize=1 filters=512 size=1 stride=1 pad=1 activation=leaky

SPP

[maxpool] stride=1 size=5

[route] layers=-2

[maxpool] stride=1 size=9

[route] layers=-4

[maxpool] stride=1 size=13

[route] layers=-1,-3,-5,-6

End SPP

[convolutional] batch_normalize=1 filters=512 size=1 stride=1 pad=1 activation=leaky

[convolutional] batch_normalize=1 size=3 stride=1 pad=1 filters=1024 activation=leaky

[convolutional] batch_normalize=1 filters=512 size=1 stride=1 pad=1 activation=leaky

[convolutional] batch_normalize=1 size=3 stride=1 pad=1 filters=1024 activation=leaky

[convolutional] size=1 stride=1 pad=1 filters=273 activation=linear

[yolo] mask = 6,7,8 anchors = 10,13, 16,30, 33,23, 30,61, 62,45, 59,119, 116,90, 156,198, 373,326 classes=86 num=9 jitter=.3 ignore_thresh = .7 truth_thresh = 1 random=1

[route] layers = -4

[convolutional] batch_normalize=1 filters=256 size=1 stride=1 pad=1 activation=leaky

[upsample] stride=4

[route] layers = -1, 11

[convolutional] batch_normalize=1 filters=256 size=1 stride=1 pad=1 activation=leaky

[convolutional] batch_normalize=1 size=3 stride=1 pad=1 filters=512 activation=leaky

[convolutional] batch_normalize=1 filters=256 size=1 stride=1 pad=1 activation=leaky

[convolutional] batch_normalize=1 size=3 stride=1 pad=1 filters=512 activation=leaky

[convolutional] batch_normalize=1 filters=256 size=1 stride=1 pad=1 activation=leaky

[convolutional] batch_normalize=1 size=3 stride=1 pad=1 filters=512 activation=leaky

[convolutional] size=1 stride=1 pad=1 filters=273 activation=linear

[yolo] mask = 3,4,5 anchors = 10,13, 16,30, 33,23, 30,61, 62,45, 59,119, 116,90, 156,198, 373,326 classes=86 num=9 jitter=.3 ignore_thresh = .7 truth_thresh = 1 random=1

[route] layers = -4

[convolutional] batch_normalize=1 filters=128 size=1 stride=1 pad=1 activation=leaky

[upsample] stride=2

[route] layers = -1, 36

[convolutional] batch_normalize=1 filters=128 size=1 stride=1 pad=1 activation=leaky

[convolutional] batch_normalize=1 size=3 stride=1 pad=1 filters=256 activation=leaky

[convolutional] batch_normalize=1 filters=128 size=1 stride=1 pad=1 activation=leaky

[convolutional] batch_normalize=1 size=3 stride=1 pad=1 filters=256 activation=leaky

[convolutional] batch_normalize=1 filters=128 size=1 stride=1 pad=1 activation=leaky

[convolutional] batch_normalize=1 size=3 stride=1 pad=1 filters=256 activation=leaky

[convolutional] size=1 stride=1 pad=1 filters=273 activation=linear

[yolo] mask = 0,1,2 anchors = 10,13, 16,30, 33,23, 30,61, 62,45, 59,119, 116,90, 156,198, 373,326 classes=86 num=9 jitter=.3 ignore_thresh = .7 truth_thresh = 1 random=1

ghost commented 4 years ago

@WongKinYiu this is my cfg file,this issue started right after i added these lines

[upsample] stride=4

[route] layers = -1, 11

WongKinYiu commented 4 years ago

the size seems not same, change

[route]
layers = -1, 11

to 

[route]
layers = -1, 12

and please give me the feedback. thanks.

~but yes, as i remember there is a bug when use [convolutional] after the [route] when routed layers are all [shoutcut], [upsample], or [maxpool]...~ after check, it already fixed several months ago.

ghost commented 4 years ago

image @WongKinYiu the error changed to this

WongKinYiu commented 4 years ago

well, i think you want to build a feature pyramid network like architecture. but you seems totally route wrong size of feature maps. please re-design the cfg and check size of feature maps carefully.

AlexeyAB commented 4 years ago

@k00r

All inputs of [route]-layer should have the same width and height: https://www.tensorflow.org/api_docs/python/tf/concat

The number of dimensions of the input tensors must match, and all dimensions except axis must be equal.

@WongKinYiu

but yes, as i remember there is a bug when use [convolutional] after the [route] when routed layers are all [shoutcut], [upsample], or [maxpool]...

There is no bug, this is just how padding=SAME works in any framework, if input 3x4, then after subsampling f.e. [maxpool] stride=2 with padding=SAME the output = 2x2, then after upsampling [upsample] stride=2 output = 4x4, so when you try to concatenate initial output 3x4 and output after subsampling+upsampling 4x4 in FPN, then you can't do it, since both inputs should be the same.

You haven’t seen this in a Pytorch simply because there is not still implemented padding=SAME https://github.com/pytorch/pytorch/issues/3867

It's 2020. Still no padding='same' in Pytorch?

With paddig=VALID you lose information around the edges, while with padding=SAME you don't lose it: https://www.pico.net/kb/what-is-the-difference-between-same-and-valid-padding-in-tf-nn-max-pool-of-tensorflow

Figure2 1581475727

TensorFlow: https://www.tensorflow.org/api_docs/python/tf/nn/convolution

If padding == "SAME": output_spatial_shape[i] = ceil(input_spatial_shape[i] / strides[i])

If padding == "VALID": output_spatial_shape[i] = ceil((input_spatial_shape[i] - (spatial_filter_shape[i]-1) * dilation_rate[i]) / strides[i]).