第1页 / 共6页
第2页 / 共6页
第3页 / 共6页
第4页 / 共6页
第5页 / 共6页
第6页 / 共6页
Yolo v3英文原文.pdf
YOLOv3: An Incremental Improvement
Joseph Redmon Ali Farhadi
University of Washington
8
1
0
2
r
p
A
8
]
V
C
.
s
c
[
1
v
7
6
7
2
0
.
4
0
8
1
:
v
i
X
r
a
Abstract
We present some updates to YOLO! We made a bunch
of little design changes to make it better. We also trained
this new network that’s pretty swell. It’s a little bigger than
last time but more accurate.
It’s still fast though, don’t
worry. At 320 × 320 YOLOv3 runs in 22 ms at 28.2 mAP,
as accurate as SSD but three times faster. When we look
at the old .5 IOU mAP detection metric YOLOv3 is quite
good. It achieves 57.9 AP50 in 51 ms on a Titan X, com-
pared to 57.5 AP50 in 198 ms by RetinaNet, similar perfor-
mance but 3.8× faster. As always, all the code is online at
https://pjreddie.com/yolo/.
1. Introduction
Sometimes you just kinda phone it in for a year, you
know? I didn’t do a whole lot of research this year. Spent
a lot of time on Twitter. Played around with GANs a little.
I had a little momentum left over from last year [12] [1]; I
managed to make some improvements to YOLO. But, hon-
estly, nothing like super interesting, just a bunch of small
changes that make it better.
I also helped out with other
people’s research a little.
Actually, that’s what brings us here today. We have
a camera-ready deadline [4] and we need to cite some of
the random updates I made to YOLO but we don’t have a
source. So get ready for a TECH REPORT!
The great thing about tech reports is that they don’t need
intros, y’all know why we’re here. So the end of this intro-
duction will signpost for the rest of the paper. First we’ll tell
you what the deal is with YOLOv3. Then we’ll tell you how
we do. We’ll also tell you about some things we tried that
didn’t work. Finally we’ll contemplate what this all means.
2. The Deal
So here’s the deal with YOLOv3: We mostly took good
ideas from other people. We also trained a new classifier
network that’s better than the other ones. We’ll just take
you through the whole system from scratch so you can un-
derstand it all.
Figure 1. We adapt this figure from the Focal Loss paper [9].
YOLOv3 runs significantly faster than other detection methods
with comparable performance. Times from either an M40 or Titan
X, they are basically the same GPU.
2.1. Bounding Box Prediction
Following YOLO9000 our system predicts bounding
boxes using dimension clusters as anchor boxes [15]. The
network predicts 4 coordinates for each bounding box, tx,
ty, tw, th. If the cell is offset from the top left corner of the
image by (cx, cy) and the bounding box prior has width and
height pw, ph, then the predictions correspond to:
bx = σ(tx) + cx
by = σ(ty) + cy
bw = pwetw
bh = pheth
During training we use sum of squared error loss. If the
ground truth for some coordinate prediction is ˆt* our gra-
dient is the ground truth value (computed from the ground
truth box) minus our prediction: ˆt* − t*. This ground truth
value can be easily computed by inverting the equations
above.
YOLOv3 predicts an objectness score for each bounding
box using logistic regression. This should be 1 if the bound-
ing box prior overlaps a ground truth object by more than
any other bounding box prior.
If the bounding box prior
1
50100150200250inference time (ms)283032343638COCO APBCDEFGRetinaNet-50RetinaNet-101YOLOv3Method[B] SSD321[C] DSSD321[D] R-FCN[E] SSD513[F] DSSD513[G] FPN FRCNRetinaNet-50-500RetinaNet-101-500RetinaNet-101-800YOLOv3-320YOLOv3-416YOLOv3-608mAP28.028.029.931.233.236.232.534.437.828.231.033.0time6185851251561727390198222951
Next we take the feature map from 2 layers previous and
upsample it by 2×. We also take a feature map from earlier
in the network and merge it with our upsampled features
using concatenation. This method allows us to get more
meaningful semantic information from the upsampled fea-
tures and finer-grained information from the earlier feature
map. We then add a few more convolutional layers to pro-
cess this combined feature map, and eventually predict a
similar tensor, although now twice the size.
We perform the same design one more time to predict
boxes for the final scale. Thus our predictions for the 3rd
scale benefit from all the prior computation as well as fine-
grained features from early on in the network.
We still use k-means clustering to determine our bound-
ing box priors. We just sort of chose 9 clusters and 3
scales arbitrarily and then divide up the clusters evenly
across scales. On the COCO dataset the 9 clusters were:
(10× 13), (16× 30), (33× 23), (30× 61), (62× 45), (59×
119), (116 × 90), (156 × 198), (373 × 326).
2.4. Feature Extractor
We use a new network for performing feature extraction.
Our new network is a hybrid approach between the network
used in YOLOv2, Darknet-19, and that newfangled residual
network stuff. Our network uses successive 3 × 3 and 1 × 1
convolutional layers but now has some shortcut connections
as well and is significantly larger. It has 53 convolutional
layers so we call it.... wait for it..... Darknet-53!
Table 1. Darknet-53.
Figure 2. Bounding boxes with dimension priors and location
prediction. We predict the width and height of the box as offsets
from cluster centroids. We predict the center coordinates of the
box relative to the location of filter application using a sigmoid
function. This figure blatantly self-plagiarized from [15].
is not the best but does overlap a ground truth object by
more than some threshold we ignore the prediction, follow-
ing [17]. We use the threshold of .5. Unlike [17] our system
only assigns one bounding box prior for each ground truth
object. If a bounding box prior is not assigned to a ground
truth object it incurs no loss for coordinate or class predic-
tions, only objectness.
2.2. Class Prediction
Each box predicts the classes the bounding box may con-
tain using multilabel classification. We do not use a softmax
as we have found it is unnecessary for good performance,
instead we simply use independent logistic classifiers. Dur-
ing training we use binary cross-entropy loss for the class
predictions.
This formulation helps when we move to more complex
domains like the Open Images Dataset [7]. In this dataset
there are many overlapping labels (i.e. Woman and Person).
Using a softmax imposes the assumption that each box has
exactly one class which is often not the case. A multilabel
approach better models the data.
2.3. Predictions Across Scales
YOLOv3 predicts boxes at 3 different scales. Our sys-
tem extracts features from those scales using a similar con-
cept to feature pyramid networks [8]. From our base fea-
ture extractor we add several convolutional layers. The last
of these predicts a 3-d tensor encoding bounding box, ob-
jectness, and class predictions.
In our experiments with
COCO [10] we predict 3 boxes at each scale so the tensor is
N × N × [3 ∗ (4 + 1 + 80)] for the 4 bounding box offsets,
1 objectness prediction, and 80 class predictions.
σ(tx)σ(ty)pwphbhbwbw=pwebh=phecxcybx=σ(tx)+cxby=σ(ty)+cytwthTypeConvolutionalConvolutionalConvolutionalConvolutionalResidualConvolutionalConvolutionalConvolutionalResidualConvolutionalConvolutionalConvolutionalResidualConvolutionalConvolutionalConvolutionalResidualConvolutionalConvolutionalConvolutionalResidualAvgpoolConnectedSoftmaxFilters326432641286412825612825651225651210245121024Size3 × 33 × 3 / 21 × 13 × 33 × 3 / 21 × 13 × 33 × 3 / 21 × 13 × 33 × 3 / 21 × 13 × 33 × 3 / 21 × 13 × 3Global1000Output256 × 256128 × 128128 × 12864 × 6464 × 6432 × 3232 × 3216 × 1616 × 168 × 88 × 81×2×8×8×4×
This new network is much more powerful than Darknet-
19 but still more efficient than ResNet-101 or ResNet-152.
Here are some ImageNet results:
Backbone
Darknet-19 [15]
ResNet-101[5]
ResNet-152 [5]
Darknet-53
Top-1
74.1
77.1
77.6
77.2
Top-5 Bn Ops BFLOP/s
1246
1039
1090
1457
91.8
93.7
93.8
93.8
7.29
19.7
29.4
18.7
FPS
171
53
37
78
Table 2. Comparison of backbones. Accuracy, billions of oper-
ations, billion floating point operations per second, and FPS for
various networks.
Each network is trained with identical settings and tested
at 256× 256, single crop accuracy. Run times are measured
on a Titan X at 256 × 256. Thus Darknet-53 performs on
par with state-of-the-art classifiers but with fewer floating
point operations and more speed. Darknet-53 is better than
ResNet-101 and 1.5× faster. Darknet-53 has similar perfor-
mance to ResNet-152 and is 2× faster.
Darknet-53 also achieves the highest measured floating
point operations per second. This means the network struc-
ture better utilizes the GPU, making it more efficient to eval-
uate and thus faster. That’s mostly because ResNets have
just way too many layers and aren’t very efficient.
2.5. Training
We still train on full images with no hard negative mining
or any of that stuff. We use multi-scale training, lots of data
augmentation, batch normalization, all the standard stuff.
We use the Darknet neural network framework for training
and testing [14].
3. How We Do
YOLOv3 is pretty good! See table 3. In terms of COCOs
weird average mean AP metric it is on par with the SSD
variants but is 3× faster. It is still quite a bit behind other
models like RetinaNet in this metric though.
However, when we look at the “old” detection metric of
mAP at IOU= .5 (or AP50 in the chart) YOLOv3 is very
strong.
It is almost on par with RetinaNet and far above
the SSD variants. This indicates that YOLOv3 is a very
strong detector that excels at producing decent boxes for ob-
jects. However, performance drops significantly as the IOU
threshold increases indicating YOLOv3 struggles to get the
boxes perfectly aligned with the object.
In the past YOLO struggled with small objects. How-
ever, now we see a reversal in that trend. With the new
multi-scale predictions we see YOLOv3 has relatively high
APS performance. However, it has comparatively worse
performance on medium and larger size objects. More in-
vestigation is needed to get to the bottom of this.
When we plot accuracy vs speed on the AP50 metric (see
figure 5) we see YOLOv3 has significant benefits over other
detection systems. Namely, it’s faster and better.
4. Things We Tried That Didn’t Work
We tried lots of stuff while we were working on
YOLOv3. A lot of it didn’t work. Here’s the stuff we can
remember.
Anchor box x, y offset predictions. We tried using the
normal anchor box prediction mechanism where you pre-
dict the x, y offset as a multiple of the box width or height
using a linear activation. We found this formulation de-
creased model stability and didn’t work very well.
Linear x, y predictions instead of logistic. We tried
using a linear activation to directly predict the x, y offset
instead of the logistic activation. This led to a couple point
drop in mAP.
Focal loss. We tried using focal loss.
It dropped our
mAP about 2 points. YOLOv3 may already be robust to
the problem focal loss is trying to solve because it has sep-
arate objectness predictions and conditional class predic-
tions. Thus for most examples there is no loss from the
class predictions? Or something? We aren’t totally sure.
Two-stage methods
Faster R-CNN+++ [5]
Faster R-CNN w FPN [8]
Faster R-CNN by G-RMI [6]
Faster R-CNN w TDM [20]
One-stage methods
YOLOv2 [15]
SSD513 [11, 3]
DSSD513 [3]
RetinaNet [9]
RetinaNet [9]
YOLOv3 608 × 608
backbone
AP
AP50
AP75
APS
APM
APL
ResNet-101-C4
ResNet-101-FPN
34.9
36.2
Inception-ResNet-v2 [21]
34.7
Inception-ResNet-v2-TDM 36.8
DarkNet-19 [15]
ResNet-101-SSD
ResNet-101-DSSD
ResNet-101-FPN
ResNeXt-101-FPN
Darknet-53
21.6
31.2
33.2
39.1
40.8
33.0
55.7
59.1
55.5
57.7
44.0
50.4
53.3
59.1
61.1
57.9
37.4
39.0
36.7
39.2
19.2
33.3
35.2
42.3
44.1
34.4
15.6
18.2
13.5
16.2
5.0
10.2
13.0
21.8
24.1
18.3
38.7
39.0
38.1
39.8
22.4
34.5
35.4
42.7
44.2
35.4
50.9
48.2
52.0
52.1
35.5
49.8
51.1
50.2
51.2
41.9
Table 3. I’m seriously just stealing all these tables from [9] they take soooo long to make from scratch. Ok, YOLOv3 is doing alright.
Keep in mind that RetinaNet has like 3.8× longer to process an image. YOLOv3 is much better than SSD variants and comparable to
state-of-the-art models on the AP50 metric.
Figure 3. Again adapted from the [9], this time displaying speed/accuracy tradeoff on the mAP at .5 IOU metric. You can tell YOLOv3 is
good because it’s very high and far to the left. Can you cite your own paper? Guess who’s going to try, this guy → [16]. Oh, I forgot, we
also fix a data loading bug in YOLOv2, that helped by like 2 mAP. Just sneaking this in here to not throw off layout.
Dual IOU thresholds and truth assignment. Faster R-
CNN uses two IOU thresholds during training. If a predic-
tion overlaps the ground truth by .7 it is as a positive exam-
ple, by [.3− .7] it is ignored, less than .3 for all ground truth
objects it is a negative example. We tried a similar strategy
but couldn’t get good results.
We quite like our current formulation, it seems to be at
a local optima at least.
It is possible that some of these
techniques could eventually produce good results, perhaps
they just need some tuning to stabilize the training.
5. What This All Means
YOLOv3 is a good detector. It’s fast, it’s accurate. It’s
not as great on the COCO average AP between .5 and .95
IOU metric. But it’s very good on the old detection metric
of .5 IOU.
Why did we switch metrics anyway? The original
COCO paper just has this cryptic sentence: “A full discus-
sion of evaluation metrics will be added once the evaluation
server is complete”. Russakovsky et al report that that hu-
mans have a hard time distinguishing an IOU of .3 from .5!
“Training humans to visually inspect a bounding box with
IOU of 0.3 and distinguish it from one with IOU 0.5 is sur-
prisingly difficult.” [18] If humans have a hard time telling
the difference, how much does it matter?
But maybe a better question is: “What are we going to
do with these detectors now that we have them?” A lot of
the people doing this research are at Google and Facebook.
I guess at least we know the technology is in good hands
and definitely won’t be used to harvest your personal infor-
mation and sell it to.... wait, you’re saying that’s exactly
what it will be used for?? Oh.
Well the other people heavily funding vision research are
the military and they’ve never done anything horrible like
killing lots of people with new technology oh wait.....1
I have a lot of hope that most of the people using com-
puter vision are just doing happy, good stuff with it, like
counting the number of zebras in a national park [13], or
tracking their cat as it wanders around their house [19]. But
computer vision is already being put to questionable use and
as researchers we have a responsibility to at least consider
the harm our work might be doing and think of ways to mit-
igate it. We owe the world that much.
In closing, do not @ me. (Because I finally quit Twitter).
1The author is funded by the Office of Naval Research and Google.
50100150200250inference time (ms)485052545658COCO mAP-50BCDEFGRetinaNet-50RetinaNet-101YOLOv3Method[B] SSD321[C] DSSD321[D] R-FCN[E] SSD513[F] DSSD513[G] FPN FRCNRetinaNet-50-500RetinaNet-101-500RetinaNet-101-800YOLOv3-320YOLOv3-416YOLOv3-608mAP-5045.446.151.950.453.359.150.953.157.551.555.357.9time6185851251561727390198222951
[18] O. Russakovsky, L.-J. Li, and L. Fei-Fei. Best of both
worlds: human-machine collaboration for object annotation.
In Proceedings of the IEEE Conference on Computer Vision
and Pattern Recognition, pages 2121–2131, 2015. 4
[19] M. Scott. Smart camera gimbal bot scanlime:027, Dec 2017.
4
[20] A. Shrivastava, R. Sukthankar, J. Malik, and A. Gupta. Be-
yond skip connections: Top-down modulation for object de-
tection. arXiv preprint arXiv:1612.06851, 2016. 3
[21] C. Szegedy, S. Ioffe, V. Vanhoucke, and A. A. Alemi.
Inception-v4, inception-resnet and the impact of residual
connections on learning. 2017. 3
References
[1] Analogy. Wikipedia, Mar 2018. 1
[2] M. Everingham, L. Van Gool, C. K. Williams, J. Winn, and
A. Zisserman. The pascal visual object classes (voc) chal-
lenge. International journal of computer vision, 88(2):303–
338, 2010. 6
[3] C.-Y. Fu, W. Liu, A. Ranga, A. Tyagi, and A. C. Berg.
Dssd: Deconvolutional single shot detector. arXiv preprint
arXiv:1701.06659, 2017. 3
[4] D. Gordon, A. Kembhavi, M. Rastegari, J. Redmon, D. Fox,
and A. Farhadi. Iqa: Visual question answering in interactive
environments. arXiv preprint arXiv:1712.03316, 2017. 1
[5] K. He, X. Zhang, S. Ren, and J. Sun. Deep residual learn-
ing for image recognition. In Proceedings of the IEEE con-
ference on computer vision and pattern recognition, pages
770–778, 2016. 3
[6] J. Huang, V. Rathod, C. Sun, M. Zhu, A. Korattikara,
A. Fathi, I. Fischer, Z. Wojna, Y. Song, S. Guadarrama, et al.
Speed/accuracy trade-offs for modern convolutional object
detectors. 3
[7] I. Krasin, T. Duerig, N. Alldrin, V. Ferrari, S. Abu-El-Haija,
A. Kuznetsova, H. Rom, J. Uijlings, S. Popov, A. Veit,
S. Belongie, V. Gomes, A. Gupta, C. Sun, G. Chechik,
D. Cai, Z. Feng, D. Narayanan, and K. Murphy. Open-
images: A public dataset for large-scale multi-label and
multi-class image classification. Dataset available from
https://github.com/openimages, 2017. 2
[8] T.-Y. Lin, P. Dollar, R. Girshick, K. He, B. Hariharan, and
S. Belongie. Feature pyramid networks for object detection.
In Proceedings of the IEEE Conference on Computer Vision
and Pattern Recognition, pages 2117–2125, 2017. 2, 3
[9] T.-Y. Lin, P. Goyal, R. Girshick, K. He, and P. Doll´ar.
arXiv preprint
loss for dense object detection.
Focal
arXiv:1708.02002, 2017. 1, 3, 4
[10] T.-Y. Lin, M. Maire, S. Belongie, J. Hays, P. Perona, D. Ra-
manan, P. Doll´ar, and C. L. Zitnick. Microsoft coco: Com-
mon objects in context. In European conference on computer
vision, pages 740–755. Springer, 2014. 2
[11] W. Liu, D. Anguelov, D. Erhan, C. Szegedy, S. Reed, C.-
Y. Fu, and A. C. Berg. Ssd: Single shot multibox detector.
In European conference on computer vision, pages 21–37.
Springer, 2016. 3
[12] I. Newton. Philosophiae naturalis principia mathematica.
William Dawson & Sons Ltd., London, 1687. 1
[13] J. Parham, J. Crall, C. Stewart, T. Berger-Wolf, and
D. Rubenstein. Animal population censusing at scale with
citizen science and photographic identification. 2017. 4
[14] J. Redmon. Darknet: Open source neural networks in c.
http://pjreddie.com/darknet/, 2013–2016. 3
[15] J. Redmon and A. Farhadi. Yolo9000: Better, faster, stronger.
In Computer Vision and Pattern Recognition (CVPR), 2017
IEEE Conference on, pages 6517–6525. IEEE, 2017. 1, 2, 3
[16] J. Redmon and A. Farhadi. Yolov3: An incremental improve-
ment. arXiv, 2018. 4
[17] S. Ren, K. He, R. Girshick, and J. Sun. Faster r-cnn: To-
wards real-time object detection with region proposal net-
works. arXiv preprint arXiv:1506.01497, 2015. 2
Figure 4. Zero-axis charts are probably more intellectually honest... and we can still screw with the variables to make ourselves look good!
Rebuttal
We would like to thank the Reddit commenters, labmates,
emailers, and passing shouts in the hallway for their lovely, heart-
felt words. If you, like me, are reviewing for ICCV then we know
you probably have 37 other papers you could be reading that you’ll
invariably put off until the last week and then have some legend in
the field email you about how you really should finish those re-
views execept it won’t entirely be clear what they’re saying and
maybe they’re from the future? Anyway, this paper won’t have be-
come what it will in time be without all the work your past selves
will have done also in the past but only a little bit further forward,
not like all the way until now forward. And if you tweeted about
it I wouldn’t know. Just sayin.
Reviewer #2 AKA Dan Grossman (lol blinding who does that)
insists that I point out here that our graphs have not one but two
non-zero origins. You’re absolutely right Dan, that’s because it
looks way better than admitting to ourselves that we’re all just
here battling over 2-3% mAP. But here are the requested graphs.
I threw in one with FPS too because we look just like super good
when we plot on FPS.
Reviewer #4 AKA JudasAdventus on Reddit writes “Entertain-
ing read but the arguments against the MSCOCO metrics seem a
bit weak”. Well, I always knew you would be the one to turn on
me Judas. You know how when you work on a project and it only
comes out alright so you have to figure out some way to justify
how what you did actually was pretty cool? I was basically trying
to do that and I lashed out at the COCO metrics a little bit. But
now that I’ve staked out this hill I may as well die on it.
See here’s the thing, mAP is already sort of broken so an up-
date to it should maybe address some of the issues with it or at least
justify why the updated version is better in some way. And that’s
the big thing I took issue with was the lack of justification. For
PASCAL VOC, the IOU threshold was ”set deliberately low to ac-
count for inaccuracies in bounding boxes in the ground truth data“
[2]. Does COCO have better labelling than VOC? This is defi-
nitely possible since COCO has segmentation masks maybe the
labels are more trustworthy and thus we aren’t as worried about
inaccuracy. But again, my problem was the lack of justification.
The COCO metric emphasizes better bounding boxes but that
emphasis must mean it de-emphasizes something else, in this case
classification accuracy. Is there a good reason to think that more
precise bounding boxes are more important than better classifi-
cation? A miss-classified example is much more obvious than a
bounding box that is slightly shifted.
mAP is already screwed up because all that matters is per-class
rank ordering. For example, if your test set only has these two
images then according to mAP two detectors that produce these
results are JUST AS GOOD:
Figure 5. These two hypothetical detectors are perfect according to
mAP over these two images. They are both perfect. Totally equal.
Now this is OBVIOUSLY an over-exaggeration of the prob-
lems with mAP but I guess my newly retconned point is that there
are such obvious discrepancies between what people in the “real
world” would care about and our current metrics that I think if
we’re going to come up with new metrics we should focus on
these discrepancies. Also, like, it’s already mean average preci-
sion, what do we even call the COCO metric, average mean aver-
age precision?
Here’s a proposal, what people actually care about is given an
image and a detector, how well will the detector find and classify
objects in the image. What about getting rid of the per-class AP
and just doing a global average precision? Or doing an AP calcu-
lation per-image and averaging over that?
Boxes are stupid anyway though, I’m probably a true believer
in masks except I can’t get YOLO to learn them.
0255075100050100150200YOLOv3All the other slow onesmAP 50Execution time (ms)0255075100012.52537.550YOLOv3All the other slow onesFPSmAP 50Person: 99%Dog: 99%Camel: 99%Bird: 99%Person: 99%Horse: 99%Detector #1Horse: 52%Person: 42%Dog: 48%Camel: 10%Bird: 90%Person: 11%Horse: 70%Detector #2Bird: 89%Horse: 60%Bird: 75%Dog: 45%
相关推荐
2023年江西萍乡中考道德与法治真题及答案.doc
2012年重庆南川中考生物真题及答案.doc
2013年江西师范大学地理学综合及文艺理论基础考研真题.doc
2020年四川甘孜小升初语文真题及答案I卷.doc
2020年注册岩土工程师专业基础考试真题及答案.doc
2023-2024学年福建省厦门市九年级上学期数学月考试题及答案.doc
2021-2022学年辽宁省沈阳市大东区九年级上学期语文期末试题及答案.doc
2022-2023学年北京东城区初三第一学期物理期末试卷及答案.doc
2018上半年江西教师资格初中地理学科知识与教学能力真题及答案.doc
2012年河北国家公务员申论考试真题及答案-省级.doc
2020-2021学年江苏省扬州市江都区邵樊片九年级上学期数学第一次质量检测试题及答案.doc
2022下半年黑龙江教师资格证中学综合素质真题及答案.doc
