Upgrade to Pro
— share decks privately, control downloads, hide ads and more …
Speaker Deck
Features
Speaker Deck
PRO
Sign in
Sign up for free
Search
Search
Deep Learning based object Detection with YOLO v2
Search
Jumabek Alikhanov
September 29, 2019
Research
1
230
Deep Learning based object Detection with YOLO v2
I will briefly go through the the process of YOLOv2
Jumabek Alikhanov
September 29, 2019
Tweet
Share
Other Decks in Research
See All in Research
2025/7/5 応用音響研究会招待講演@北海道大学
takuma_okamoto
1
190
AWSで実現した大規模日本語VLM学習用データセット "MOMIJI" 構築パイプライン/buiding-momiji
studio_graph
2
520
Minimax and Bayes Optimal Best-arm Identification: Adaptive Experimental Design for Treatment Choice
masakat0
0
170
ストレス計測方法の確立に向けたマルチモーダルデータの活用
yurikomium
0
1.5k
単施設でできる臨床研究の考え方
shuntaros
0
2.7k
[輪講] SigLIP 2: Multilingual Vision-Language Encoders with Improved Semantic Understanding, Localization, and Dense Features
nk35jk
2
1k
数理最適化と機械学習の融合
mickey_kubo
16
9.3k
不確実性下における目的と手段の統合的探索に向けた連続腕バンディットの応用 / iot70_gp_rff_mab
monochromegane
2
170
電通総研の生成AI・エージェントの取り組みエンジニアリング業務向けAI活用事例紹介
isidaitc
1
960
最適化と機械学習による問題解決
mickey_kubo
0
170
とあるSREの博士「過程」 / A Certain SRE’s Ph.D. Journey
yuukit
10
4.2k
MetaEarth: A Generative Foundation Model for Global-Scale Remote Sensing Image Generation
satai
4
200
Featured
See All Featured
Cheating the UX When There Is Nothing More to Optimize - PixelPioneers
stephaniewalter
285
14k
Imperfection Machines: The Place of Print at Facebook
scottboms
268
13k
Dealing with People You Can't Stand - Big Design 2015
cassininazir
367
27k
Understanding Cognitive Biases in Performance Measurement
bluesmoon
29
1.9k
JavaScript: Past, Present, and Future - NDC Porto 2020
reverentgeek
52
5.6k
Side Projects
sachag
455
43k
Measuring & Analyzing Core Web Vitals
bluesmoon
9
590
Build The Right Thing And Hit Your Dates
maggiecrowley
37
2.9k
個人開発の失敗を避けるイケてる考え方 / tips for indie hackers
panda_program
113
20k
Why Our Code Smells
bkeepers
PRO
339
57k
Put a Button on it: Removing Barriers to Going Fast.
kastner
60
4k
[Rails World 2023 - Day 1 Closing Keynote] - The Magic of Rails
eileencodes
36
2.5k
Transcript
Jumabek Alikhanov @Information Security Research Lab, Inha University YOLO9000: Better,
Faster, Stronger (CVPR 2017, Best Paper Honorable Mention) 1
1. Introduction & Previous Work 2. Better detection performance 3.
Faster processing speed 4. Detecting more classes(object types) 5. Conclusion CONTENTS 2
Task & Evaluation Metric mAP- mean Avarage Precision 3 https://github.com/rafaelpadilla/Object-Detection-Metrics
YOLO v1 Network Output shape = (S, S, B×5 +
C) = (7, 7, 2×5 + 20) = (7, 7, 30). 4
YOLOv1: Loss Function pi-conditional class Prob. Ci - box confidence
score 5 Localization Confidence Classification
Previously Pascal 2007 mAP Speed DPM v5 33.7 .07 FPS
14 s/img R-CNN 66.0 .05 FPS 20 s/img Fast R-CNN 70.0 .5 FPS 2 s/img Faster R-CNN 73.2 7 FPS 140 ms/img YOLO 63.4 45 FPS 22 ms/img 6
Previously Pascal 2007 mAP Speed DPM v5 33.7 .07 FPS
14 s/img R-CNN 66.0 .05 FPS 20 s/img Fast R-CNN 70.0 .5 FPS 2 s/img Faster R-CNN 73.2 7 FPS 140 ms/img YOLO 63.4 45 FPS 22 ms/img 7
Better Performance 8
9 YOLO Train on ImageNet Fine-tune on detection Resize network
10 Fine-tune 448x448 Classifier: +3.5% mAP Train on ImageNet Fine-tune
on detection Resize, fine-tune on ImageNet
Anchor boxes use static initialization
Use k-means clustering to find better initializations https://github.com/Jumabek/darknet_scripts
None
Static Anchors vs Dimension Clusters 14
Box Location Prediction 15
Dimension Clusters: +5% mAP
17 Multi-scale training: +1.5% mAP
YOLOv2: Fast, Accurate Detection
Huang, Jonathan, et al. "Speed/accuracy trade-offs for modern convolutional object
detectors." arXiv preprint arXiv:1611.10012 (2016).
Huang, Jonathan, et al. "Speed/accuracy trade-offs for modern convolutional object
detectors." arXiv preprint arXiv:1611.10012 (2016).
Huang, Jonathan, et al. "Speed/accuracy trade-offs for modern convolutional object
detectors." arXiv preprint arXiv:1611.10012 (2016). YOLOv2
None
Faster Detection Speed 23
Speed is not just parameter counts or FLOPs Top 1
Top 5 FLOPs GPU Speed VGG-16 70.5 90.0 30.95 Bn 100 FPS Extraction (YOLOv1) 72.5 90.8 8.52 Bn 180 FPS Resnet50 75.3 92.2 7.66 Bn 90 FPS
Darknet19: A good balance of speed and accuracy Top 1
Top 5 FLOPs GPU Speed VGG-16 70.5 90.0 30.95 Bn 100 FPS Extraction (YOLOv1) 72.5 90.8 8.52 Bn 180 FPS Resnet50 75.3 92.2 7.66 Bn 90 FPS Darknet19 74.0 91.8 5.58 Bn 200 FPS
Why is it fast? Simple & efficient architecture C implementation
26
Stronger - Detecting more classes 27
- 14 million images - 22k classes - Classification labels
- 100k images - 80 classes - Detection labels Golden eagle
Typically use softmax over all classes
Can’t just mash classes together...
Can’t just mash classes together...
WordNet has structure but it’s messy
None
None
... Each node is a conditional probability
... Each node is a conditional probability P(Bedlington terrier) =
P(object) * P (living thing | object) * ….. P(canine | mammal) * P(dog | canine) * P(terrier | dog) * P(Bedlington terrier | terrier)
None
None
None
None
None
None
None
None
None
None
Conclusion • YOLOv2 and YOLO9000 real-time detection systems • YOLOv2
state of the art and faster than other systems • 9K object category detection by YOLO9000 47
1. CVPR paper - https://pjreddie.com/media/files/papers/YOLO9000.pdf 2. Article - https://medium.com/@jonathan_hui/real-time-object-detection-with-yolo-yolov2-28b1b93e2088 3.
Author’s Presentation - https://docs.google.com/presentation/d/14qBAiyhMOFl_wZW4dA1CkixgXwf0zKGbpw_0oHK8yEM/edit#slide=id.g1f9fb98e4b_0 _132 References 48