SlideShare a Scribd company logo
1 of 21
Kaggle ultrasound nerve segmentation
Tyantov Eduard
#1 Description
Thu 19 May 2016 – Thu 18 Aug 2016
#2 Data: to find Brachial Plexus (BP)
– 420x580 resolution
– 5635 train images with masks, 5508 test;
– ~120 images per 47 patients
– 47% of the images don’t have a mask;
– result in RLE encoding
#3 Data: mistakes in the ground truth
– 45 known errors of near duplicate images
– metric is sensitive to nerve presence mistakes
#4 Evaluation
Peculiarities
– (!) Mask presence mistake leads to zero score
– Needs smoothing in the denominator
Loss functions
– 1 – dice
– -dice
– weighted cross entropy (2 classes, per pixel prediction)
Mean mask
#5 Baselines
Score Description Framework Author
0.51 Empty submission - -
0.00 Top left pixel - -
0.57 U-Net, in the
beginning of the
competition
Keras code Marko Jocic, kaggler
0.62 U-Net, almost at the
end
Torch code Qure.ai (host)
#6 What is U-Net
Overview
– (May 2015 article) “U-Net: Convolutional Networks for Biomedical Image Segmentation”
– Winner of “Grand Challenge for Computer-Automated Detection of Caries in Bitewing
Radiography at ISBI 2015”
– Encoder-decoder architecture with skip connection on the same level
– Fully convolutional, Drop-out in the middle
– Augmentation: “Smooth deformations using random displacement vectors on a coarse 3 by 3
grid. The displacements are sampled from a Gaussian distribution with 10 pixels standard
deviation.”
#7 Another aproach, FCN
Overview
– (20 May 2016, article) “Fully Convolutional Networks for Semantic Segmentation”
– VGG-18
– Segmentation prediction on different layers of the net, +upsampling
– Average predictions
#8 Starting point, Marko Jocic’s solution
Overview
– Classic U-Net: VGG-like
– Very simple Keras code
– Image resize to 64x80, bicubic interpolation
– Loss= - Dice coefficient, per batch averaging, smooth=1
– Training on whole dataset, no validation
– RLE-encoding function
– Adam optimizer
Training
– 20 epochs, ~30 seconds on Titan X, memory footprint 800mb
– Overfits, 0.68 on training -> 0.57 on LB
#9 Aspects of the solution: basics
Overfitting basics (+2%)
– Split train/valid, 20% and early stopping patience=5 epochs
• used random split instead of more convenient by patient (due to a subtle bug)
– Dropout after each conv layer
General enhancements
– Resolution 64x80 -> 80x112 (+1%)
– ELU instead of ReLU -> faster convergence
#10 Aspects of the solution: augmentation
Augmentation*
– flip x, y
– random rotate (5)
– random zoom (0.9, 1.1)
– random channel shift (5.0)
*all transformations should be done with a mask too
All transformations can be done on the fly with a generator (randomly applied), but didn’t improve
results.
Elastic transform:
convolve with a Gaussian on random
displacement fields
Result: no added effect
#11 Aspects of the solution: blocks
Modifications of U-Net
– 2 3x3 convolution -> inception_v3 block
– BNA after each convolution
– BNA + activation after summation
– nxn -> 1xn + nx1
Results:
– lesser parameters (1M)
– faster convergence
– LB: +2%
v3 block
v3 + splitted
#12 Aspects of the solution: 2nd head, postfilter
2nd head
– mask presence branch in the middle of NN (after decoder part)
• Conv 1x1, sigmoid
• FC=1, sigmoid
– leads to better convergence
Post filter
– presence prob < 0.5 or sum(pixels) < 3000 -> empty mask (+4.5%)
– in the end: combining p_nerve = (p_score + p_segment)/2 -> +0.5%
#13 Aspects of the solution: other
Modifications
– Skip connection with Residual blocks (+1%)
– Max pool -> Conv 3x3 with stride=2 (+BNA)
– Ensemble (+1%)
• k-fold 5,6,7,8, average
– Prediction on augmented versions of test images (averaging)
Final result:
– single model 0.694 score
– ensemble 0.70399 (hour before the competition’s end)
– last submission has been human verified ;) but no help
code: https://github.com/EdwardTyantov/ultrasound-nerve-segmentation
#14 Leaderboard
– 31-th
– see Private LB huge shake-up
#15 What didn’t help
– Inception Resnet v4
– sequential training of decoder, encoder parts
– more or lesser layers/blocks/n_filters
– pixel clustering
– higher or lower resolution
– dropout, different probs
– Torch version
– deconv layers instead of upsampling
– weight decay for layers
– FCN
– Deepmask architecture
#16 Technical
– Ubuntu 14 or 16, Cuda 8, cudnn 5, keras last, torch last
– batch_size=64, 128 (depends on GPU memory)
– Single model, 2-3 hours on Titan/1080
– Ensemble – 24 hours
– train dataset: error re-labeling or zero-outing
– FCN with several heads in different resolutions (regularize)
– post process: mask to elipse, no holes
– separate training: mask/no mask
– crop images, super-resolution
– models on different resolution
– higher resolution
– loss
– smart post-processing
• which obv. led to overfitting on Public Score
– replication padding instead of zero-padding
#17 Other competitors
#18 Deepmask (FB)
– no low-level features
– CNN with two heads: mask and score
– training set: patch, mask, y_k - objected in centered and fully contained
• mask pixel=1 if it is part of object in the center
– VGG, 8-layers, can be trained
– Training
• joint learning score * 1/32
• first branch only positives
• augmentation: shift 16pix, scale little, horiz. flip
– Evaluation
• full image: 16 pixel stride
#19 Sharpmask (refined deepmask)
#20 Sharpmask (refined deepmask)
– Architecture
• trunk: resnet-50 pretrained
• reflect-pad instead of zero
• for DeepMask: headC
– Training
• Deepmask (score + coarse mask)
• then freeze and learn refinements
• Why: faster convergenc, deepmask or sharpmask result, minimal finetuning
– Inference
• only top-n locations refined

More Related Content

What's hot

random forest regression
random forest regressionrandom forest regression
random forest regressionAkhilesh Joshi
 
Lect5 principal component analysis
Lect5 principal component analysisLect5 principal component analysis
Lect5 principal component analysishktripathy
 
Understanding your data with Bayesian networks (in Python) by Bartek Wilczyns...
Understanding your data with Bayesian networks (in Python) by Bartek Wilczyns...Understanding your data with Bayesian networks (in Python) by Bartek Wilczyns...
Understanding your data with Bayesian networks (in Python) by Bartek Wilczyns...PyData
 
Support vector machines (svm)
Support vector machines (svm)Support vector machines (svm)
Support vector machines (svm)Sharayu Patil
 
Machine learning basics using trees algorithm (Random forest, Gradient Boosting)
Machine learning basics using trees algorithm (Random forest, Gradient Boosting)Machine learning basics using trees algorithm (Random forest, Gradient Boosting)
Machine learning basics using trees algorithm (Random forest, Gradient Boosting)Parth Khare
 
Data Mining: Concepts and Techniques (3rd ed.) - Chapter 3 preprocessing
Data Mining:  Concepts and Techniques (3rd ed.)- Chapter 3 preprocessingData Mining:  Concepts and Techniques (3rd ed.)- Chapter 3 preprocessing
Data Mining: Concepts and Techniques (3rd ed.) - Chapter 3 preprocessingSalah Amean
 
Image processing
Image processingImage processing
Image processingPooja G N
 
Convolutional neural network
Convolutional neural networkConvolutional neural network
Convolutional neural networkMojammilHusain
 
Random forest
Random forestRandom forest
Random forestUjjawal
 
Convolutional Neural Networks (CNN)
Convolutional Neural Networks (CNN)Convolutional Neural Networks (CNN)
Convolutional Neural Networks (CNN)Gaurav Mittal
 
ViT (Vision Transformer) Review [CDM]
ViT (Vision Transformer) Review [CDM]ViT (Vision Transformer) Review [CDM]
ViT (Vision Transformer) Review [CDM]Dongmin Choi
 
Random Forest Algorithm - Random Forest Explained | Random Forest In Machine ...
Random Forest Algorithm - Random Forest Explained | Random Forest In Machine ...Random Forest Algorithm - Random Forest Explained | Random Forest In Machine ...
Random Forest Algorithm - Random Forest Explained | Random Forest In Machine ...Simplilearn
 
Support Vector Machine ppt presentation
Support Vector Machine ppt presentationSupport Vector Machine ppt presentation
Support Vector Machine ppt presentationAyanaRukasar
 
Image Processing with OpenCV
Image Processing with OpenCVImage Processing with OpenCV
Image Processing with OpenCVdebayanin
 

What's hot (20)

random forest regression
random forest regressionrandom forest regression
random forest regression
 
Lect5 principal component analysis
Lect5 principal component analysisLect5 principal component analysis
Lect5 principal component analysis
 
PCA Final.pptx
PCA Final.pptxPCA Final.pptx
PCA Final.pptx
 
Data Augmentation
Data AugmentationData Augmentation
Data Augmentation
 
Understanding your data with Bayesian networks (in Python) by Bartek Wilczyns...
Understanding your data with Bayesian networks (in Python) by Bartek Wilczyns...Understanding your data with Bayesian networks (in Python) by Bartek Wilczyns...
Understanding your data with Bayesian networks (in Python) by Bartek Wilczyns...
 
Support vector machines (svm)
Support vector machines (svm)Support vector machines (svm)
Support vector machines (svm)
 
PCA
PCAPCA
PCA
 
Machine learning basics using trees algorithm (Random forest, Gradient Boosting)
Machine learning basics using trees algorithm (Random forest, Gradient Boosting)Machine learning basics using trees algorithm (Random forest, Gradient Boosting)
Machine learning basics using trees algorithm (Random forest, Gradient Boosting)
 
Data Mining: Concepts and Techniques (3rd ed.) - Chapter 3 preprocessing
Data Mining:  Concepts and Techniques (3rd ed.)- Chapter 3 preprocessingData Mining:  Concepts and Techniques (3rd ed.)- Chapter 3 preprocessing
Data Mining: Concepts and Techniques (3rd ed.) - Chapter 3 preprocessing
 
Image processing
Image processingImage processing
Image processing
 
Convolutional neural network
Convolutional neural networkConvolutional neural network
Convolutional neural network
 
Pca ppt
Pca pptPca ppt
Pca ppt
 
Random forest
Random forestRandom forest
Random forest
 
Convolutional Neural Networks (CNN)
Convolutional Neural Networks (CNN)Convolutional Neural Networks (CNN)
Convolutional Neural Networks (CNN)
 
ViT (Vision Transformer) Review [CDM]
ViT (Vision Transformer) Review [CDM]ViT (Vision Transformer) Review [CDM]
ViT (Vision Transformer) Review [CDM]
 
boosting algorithm
boosting algorithmboosting algorithm
boosting algorithm
 
Random Forest Algorithm - Random Forest Explained | Random Forest In Machine ...
Random Forest Algorithm - Random Forest Explained | Random Forest In Machine ...Random Forest Algorithm - Random Forest Explained | Random Forest In Machine ...
Random Forest Algorithm - Random Forest Explained | Random Forest In Machine ...
 
Support Vector Machine ppt presentation
Support Vector Machine ppt presentationSupport Vector Machine ppt presentation
Support Vector Machine ppt presentation
 
Background subtraction
Background subtractionBackground subtraction
Background subtraction
 
Image Processing with OpenCV
Image Processing with OpenCVImage Processing with OpenCV
Image Processing with OpenCV
 

Viewers also liked

Artisto App, Highload 2016
Artisto App, Highload 2016Artisto App, Highload 2016
Artisto App, Highload 2016Eduard Tyantov
 
Intro to Machine Learning by Microsoft Ventures
Intro to Machine Learning by Microsoft VenturesIntro to Machine Learning by Microsoft Ventures
Intro to Machine Learning by Microsoft Venturesmicrosoftventures
 
Scalable TensorFlow Deep Learning as a Service with Docker, OpenPOWER, and GPUs
Scalable TensorFlow Deep Learning as a Service with Docker, OpenPOWER, and GPUsScalable TensorFlow Deep Learning as a Service with Docker, OpenPOWER, and GPUs
Scalable TensorFlow Deep Learning as a Service with Docker, OpenPOWER, and GPUsIndrajit Poddar
 
Julian - diagnosing heart disease using convolutional neural networks
Julian - diagnosing heart disease using convolutional neural networksJulian - diagnosing heart disease using convolutional neural networks
Julian - diagnosing heart disease using convolutional neural networksAbhishek Thakur
 
Google Dev Summit Extended Seoul - TensorFlow: Tensorboard & Keras
Google Dev Summit Extended Seoul - TensorFlow: Tensorboard & KerasGoogle Dev Summit Extended Seoul - TensorFlow: Tensorboard & Keras
Google Dev Summit Extended Seoul - TensorFlow: Tensorboard & KerasTaegyun Jeon
 

Viewers also liked (6)

Acúfenos preevid
Acúfenos preevidAcúfenos preevid
Acúfenos preevid
 
Artisto App, Highload 2016
Artisto App, Highload 2016Artisto App, Highload 2016
Artisto App, Highload 2016
 
Intro to Machine Learning by Microsoft Ventures
Intro to Machine Learning by Microsoft VenturesIntro to Machine Learning by Microsoft Ventures
Intro to Machine Learning by Microsoft Ventures
 
Scalable TensorFlow Deep Learning as a Service with Docker, OpenPOWER, and GPUs
Scalable TensorFlow Deep Learning as a Service with Docker, OpenPOWER, and GPUsScalable TensorFlow Deep Learning as a Service with Docker, OpenPOWER, and GPUs
Scalable TensorFlow Deep Learning as a Service with Docker, OpenPOWER, and GPUs
 
Julian - diagnosing heart disease using convolutional neural networks
Julian - diagnosing heart disease using convolutional neural networksJulian - diagnosing heart disease using convolutional neural networks
Julian - diagnosing heart disease using convolutional neural networks
 
Google Dev Summit Extended Seoul - TensorFlow: Tensorboard & Keras
Google Dev Summit Extended Seoul - TensorFlow: Tensorboard & KerasGoogle Dev Summit Extended Seoul - TensorFlow: Tensorboard & Keras
Google Dev Summit Extended Seoul - TensorFlow: Tensorboard & Keras
 

Similar to Ultrasound nerve segmentation, kaggle review

Salt Identification Challenge
Salt Identification ChallengeSalt Identification Challenge
Salt Identification Challengekenluck2001
 
QGATE 0.3: QUANTUM CIRCUIT SIMULATOR
QGATE 0.3: QUANTUM CIRCUIT SIMULATORQGATE 0.3: QUANTUM CIRCUIT SIMULATOR
QGATE 0.3: QUANTUM CIRCUIT SIMULATORNVIDIA Japan
 
GPGPU Computation
GPGPU ComputationGPGPU Computation
GPGPU Computationjtsagata
 
Jeff Johnson, Research Engineer, Facebook at MLconf NYC
Jeff Johnson, Research Engineer, Facebook at MLconf NYCJeff Johnson, Research Engineer, Facebook at MLconf NYC
Jeff Johnson, Research Engineer, Facebook at MLconf NYCMLconf
 
Designing High Performance Computing Architectures for Reliable Space Applica...
Designing High Performance Computing Architectures for Reliable Space Applica...Designing High Performance Computing Architectures for Reliable Space Applica...
Designing High Performance Computing Architectures for Reliable Space Applica...Fisnik Kraja
 
Recent Progress in SCCS on GPU Simulation of Biomedical and Hydrodynamic Prob...
Recent Progress in SCCS on GPU Simulation of Biomedical and Hydrodynamic Prob...Recent Progress in SCCS on GPU Simulation of Biomedical and Hydrodynamic Prob...
Recent Progress in SCCS on GPU Simulation of Biomedical and Hydrodynamic Prob...NVIDIA Taiwan
 
Monte Carlo on GPUs
Monte Carlo on GPUsMonte Carlo on GPUs
Monte Carlo on GPUsfcassier
 
Csw2016 wheeler barksdale-gruskovnjak-execute_mypacket
Csw2016 wheeler barksdale-gruskovnjak-execute_mypacketCsw2016 wheeler barksdale-gruskovnjak-execute_mypacket
Csw2016 wheeler barksdale-gruskovnjak-execute_mypacketCanSecWest
 
Accelerating HPC Applications on NVIDIA GPUs with OpenACC
Accelerating HPC Applications on NVIDIA GPUs with OpenACCAccelerating HPC Applications on NVIDIA GPUs with OpenACC
Accelerating HPC Applications on NVIDIA GPUs with OpenACCinside-BigData.com
 
NVIDIA深度學習教育機構 (DLI): Object detection with jetson
NVIDIA深度學習教育機構 (DLI): Object detection with jetsonNVIDIA深度學習教育機構 (DLI): Object detection with jetson
NVIDIA深度學習教育機構 (DLI): Object detection with jetsonNVIDIA Taiwan
 
Face Recognition: From Scratch To Hatch
Face Recognition: From Scratch To HatchFace Recognition: From Scratch To Hatch
Face Recognition: From Scratch To HatchEduard Tyantov
 
Face Recognition: From Scratch To Hatch / Эдуард Тянтов (Mail.ru Group)
Face Recognition: From Scratch To Hatch / Эдуард Тянтов (Mail.ru Group)Face Recognition: From Scratch To Hatch / Эдуард Тянтов (Mail.ru Group)
Face Recognition: From Scratch To Hatch / Эдуард Тянтов (Mail.ru Group)Ontico
 
digitaldesign-s20-lecture3b-fpga-afterlecture.pdf
digitaldesign-s20-lecture3b-fpga-afterlecture.pdfdigitaldesign-s20-lecture3b-fpga-afterlecture.pdf
digitaldesign-s20-lecture3b-fpga-afterlecture.pdfDuy-Hieu Bui
 
Nvidia in bioinformatics
Nvidia in bioinformaticsNvidia in bioinformatics
Nvidia in bioinformaticsShanker Trivedi
 
Using ARM Dev.Board in physical experimental instruments
Using ARM Dev.Board in physical experimental instrumentsUsing ARM Dev.Board in physical experimental instruments
Using ARM Dev.Board in physical experimental instrumentsa_n0v
 
Advances in the Solution of Navier-Stokes Eqs. in GPGPU Hardware. Modelling F...
Advances in the Solution of Navier-Stokes Eqs. in GPGPU Hardware. Modelling F...Advances in the Solution of Navier-Stokes Eqs. in GPGPU Hardware. Modelling F...
Advances in the Solution of Navier-Stokes Eqs. in GPGPU Hardware. Modelling F...Storti Mario
 

Similar to Ultrasound nerve segmentation, kaggle review (20)

Salt Identification Challenge
Salt Identification ChallengeSalt Identification Challenge
Salt Identification Challenge
 
QGATE 0.3: QUANTUM CIRCUIT SIMULATOR
QGATE 0.3: QUANTUM CIRCUIT SIMULATORQGATE 0.3: QUANTUM CIRCUIT SIMULATOR
QGATE 0.3: QUANTUM CIRCUIT SIMULATOR
 
Gpu perf-presentation
Gpu perf-presentationGpu perf-presentation
Gpu perf-presentation
 
GPGPU Computation
GPGPU ComputationGPGPU Computation
GPGPU Computation
 
Eye deep
Eye deepEye deep
Eye deep
 
Jeff Johnson, Research Engineer, Facebook at MLconf NYC
Jeff Johnson, Research Engineer, Facebook at MLconf NYCJeff Johnson, Research Engineer, Facebook at MLconf NYC
Jeff Johnson, Research Engineer, Facebook at MLconf NYC
 
Designing High Performance Computing Architectures for Reliable Space Applica...
Designing High Performance Computing Architectures for Reliable Space Applica...Designing High Performance Computing Architectures for Reliable Space Applica...
Designing High Performance Computing Architectures for Reliable Space Applica...
 
Recent Progress in SCCS on GPU Simulation of Biomedical and Hydrodynamic Prob...
Recent Progress in SCCS on GPU Simulation of Biomedical and Hydrodynamic Prob...Recent Progress in SCCS on GPU Simulation of Biomedical and Hydrodynamic Prob...
Recent Progress in SCCS on GPU Simulation of Biomedical and Hydrodynamic Prob...
 
Monte Carlo on GPUs
Monte Carlo on GPUsMonte Carlo on GPUs
Monte Carlo on GPUs
 
Csw2016 wheeler barksdale-gruskovnjak-execute_mypacket
Csw2016 wheeler barksdale-gruskovnjak-execute_mypacketCsw2016 wheeler barksdale-gruskovnjak-execute_mypacket
Csw2016 wheeler barksdale-gruskovnjak-execute_mypacket
 
Accelerating HPC Applications on NVIDIA GPUs with OpenACC
Accelerating HPC Applications on NVIDIA GPUs with OpenACCAccelerating HPC Applications on NVIDIA GPUs with OpenACC
Accelerating HPC Applications on NVIDIA GPUs with OpenACC
 
NVIDIA深度學習教育機構 (DLI): Object detection with jetson
NVIDIA深度學習教育機構 (DLI): Object detection with jetsonNVIDIA深度學習教育機構 (DLI): Object detection with jetson
NVIDIA深度學習教育機構 (DLI): Object detection with jetson
 
Face Recognition: From Scratch To Hatch
Face Recognition: From Scratch To HatchFace Recognition: From Scratch To Hatch
Face Recognition: From Scratch To Hatch
 
Face Recognition: From Scratch To Hatch / Эдуард Тянтов (Mail.ru Group)
Face Recognition: From Scratch To Hatch / Эдуард Тянтов (Mail.ru Group)Face Recognition: From Scratch To Hatch / Эдуард Тянтов (Mail.ru Group)
Face Recognition: From Scratch To Hatch / Эдуард Тянтов (Mail.ru Group)
 
Can FPGAs Compete with GPUs?
Can FPGAs Compete with GPUs?Can FPGAs Compete with GPUs?
Can FPGAs Compete with GPUs?
 
digitaldesign-s20-lecture3b-fpga-afterlecture.pdf
digitaldesign-s20-lecture3b-fpga-afterlecture.pdfdigitaldesign-s20-lecture3b-fpga-afterlecture.pdf
digitaldesign-s20-lecture3b-fpga-afterlecture.pdf
 
Exploring Gpgpu Workloads
Exploring Gpgpu WorkloadsExploring Gpgpu Workloads
Exploring Gpgpu Workloads
 
Nvidia in bioinformatics
Nvidia in bioinformaticsNvidia in bioinformatics
Nvidia in bioinformatics
 
Using ARM Dev.Board in physical experimental instruments
Using ARM Dev.Board in physical experimental instrumentsUsing ARM Dev.Board in physical experimental instruments
Using ARM Dev.Board in physical experimental instruments
 
Advances in the Solution of Navier-Stokes Eqs. in GPGPU Hardware. Modelling F...
Advances in the Solution of Navier-Stokes Eqs. in GPGPU Hardware. Modelling F...Advances in the Solution of Navier-Stokes Eqs. in GPGPU Hardware. Modelling F...
Advances in the Solution of Navier-Stokes Eqs. in GPGPU Hardware. Modelling F...
 

More from Eduard Tyantov

Эксплуатация ML в Почте Mail.ru
Эксплуатация ML в Почте Mail.ruЭксплуатация ML в Почте Mail.ru
Эксплуатация ML в Почте Mail.ruEduard Tyantov
 
Опыт моделеварения от команды ComputerVision Mail.ru
Опыт моделеварения от команды ComputerVision Mail.ruОпыт моделеварения от команды ComputerVision Mail.ru
Опыт моделеварения от команды ComputerVision Mail.ruEduard Tyantov
 
Саморазвитие: как я не усидел на двух стульях и нашел третий
Саморазвитие: как я не усидел на двух стульях и нашел третийСаморазвитие: как я не усидел на двух стульях и нашел третий
Саморазвитие: как я не усидел на двух стульях и нашел третийEduard Tyantov
 
Project Management 2.0: AI Transformation
Project Management 2.0: AI TransformationProject Management 2.0: AI Transformation
Project Management 2.0: AI TransformationEduard Tyantov
 
Kaggle Google Landmark recognition
Kaggle Google Landmark recognitionKaggle Google Landmark recognition
Kaggle Google Landmark recognitionEduard Tyantov
 
Deep Learning: Advances Of The Last Year
Deep Learning: Advances Of The Last Year Deep Learning: Advances Of The Last Year
Deep Learning: Advances Of The Last Year Eduard Tyantov
 
Kaggle review Planet: Understanding the Amazon from Space
Kaggle reviewPlanet: Understanding the Amazon from SpaceKaggle reviewPlanet: Understanding the Amazon from Space
Kaggle review Planet: Understanding the Amazon from SpaceEduard Tyantov
 

More from Eduard Tyantov (7)

Эксплуатация ML в Почте Mail.ru
Эксплуатация ML в Почте Mail.ruЭксплуатация ML в Почте Mail.ru
Эксплуатация ML в Почте Mail.ru
 
Опыт моделеварения от команды ComputerVision Mail.ru
Опыт моделеварения от команды ComputerVision Mail.ruОпыт моделеварения от команды ComputerVision Mail.ru
Опыт моделеварения от команды ComputerVision Mail.ru
 
Саморазвитие: как я не усидел на двух стульях и нашел третий
Саморазвитие: как я не усидел на двух стульях и нашел третийСаморазвитие: как я не усидел на двух стульях и нашел третий
Саморазвитие: как я не усидел на двух стульях и нашел третий
 
Project Management 2.0: AI Transformation
Project Management 2.0: AI TransformationProject Management 2.0: AI Transformation
Project Management 2.0: AI Transformation
 
Kaggle Google Landmark recognition
Kaggle Google Landmark recognitionKaggle Google Landmark recognition
Kaggle Google Landmark recognition
 
Deep Learning: Advances Of The Last Year
Deep Learning: Advances Of The Last Year Deep Learning: Advances Of The Last Year
Deep Learning: Advances Of The Last Year
 
Kaggle review Planet: Understanding the Amazon from Space
Kaggle reviewPlanet: Understanding the Amazon from SpaceKaggle reviewPlanet: Understanding the Amazon from Space
Kaggle review Planet: Understanding the Amazon from Space
 

Recently uploaded

Call Girls Service Jaipur Grishma WhatsApp ❤8445551418 VIP Call Girls Jaipur
Call Girls Service Jaipur Grishma WhatsApp ❤8445551418 VIP Call Girls JaipurCall Girls Service Jaipur Grishma WhatsApp ❤8445551418 VIP Call Girls Jaipur
Call Girls Service Jaipur Grishma WhatsApp ❤8445551418 VIP Call Girls Jaipurparulsinha
 
Top Rated Bangalore Call Girls Richmond Circle ⟟ 8250192130 ⟟ Call Me For Gen...
Top Rated Bangalore Call Girls Richmond Circle ⟟ 8250192130 ⟟ Call Me For Gen...Top Rated Bangalore Call Girls Richmond Circle ⟟ 8250192130 ⟟ Call Me For Gen...
Top Rated Bangalore Call Girls Richmond Circle ⟟ 8250192130 ⟟ Call Me For Gen...narwatsonia7
 
Call Girls Kochi Just Call 9907093804 Top Class Call Girl Service Available
Call Girls Kochi Just Call 9907093804 Top Class Call Girl Service AvailableCall Girls Kochi Just Call 9907093804 Top Class Call Girl Service Available
Call Girls Kochi Just Call 9907093804 Top Class Call Girl Service AvailableDipal Arora
 
Call Girls Aurangabad Just Call 9907093804 Top Class Call Girl Service Available
Call Girls Aurangabad Just Call 9907093804 Top Class Call Girl Service AvailableCall Girls Aurangabad Just Call 9907093804 Top Class Call Girl Service Available
Call Girls Aurangabad Just Call 9907093804 Top Class Call Girl Service AvailableDipal Arora
 
Chandrapur Call girls 8617370543 Provides all area service COD available
Chandrapur Call girls 8617370543 Provides all area service COD availableChandrapur Call girls 8617370543 Provides all area service COD available
Chandrapur Call girls 8617370543 Provides all area service COD availableDipal Arora
 
Call Girls Bhubaneswar Just Call 9907093804 Top Class Call Girl Service Avail...
Call Girls Bhubaneswar Just Call 9907093804 Top Class Call Girl Service Avail...Call Girls Bhubaneswar Just Call 9907093804 Top Class Call Girl Service Avail...
Call Girls Bhubaneswar Just Call 9907093804 Top Class Call Girl Service Avail...Dipal Arora
 
(Rocky) Jaipur Call Girl - 09521753030 Escorts Service 50% Off with Cash ON D...
(Rocky) Jaipur Call Girl - 09521753030 Escorts Service 50% Off with Cash ON D...(Rocky) Jaipur Call Girl - 09521753030 Escorts Service 50% Off with Cash ON D...
(Rocky) Jaipur Call Girl - 09521753030 Escorts Service 50% Off with Cash ON D...indiancallgirl4rent
 
VIP Service Call Girls Sindhi Colony 📳 7877925207 For 18+ VIP Call Girl At Th...
VIP Service Call Girls Sindhi Colony 📳 7877925207 For 18+ VIP Call Girl At Th...VIP Service Call Girls Sindhi Colony 📳 7877925207 For 18+ VIP Call Girl At Th...
VIP Service Call Girls Sindhi Colony 📳 7877925207 For 18+ VIP Call Girl At Th...jageshsingh5554
 
Call Girls Horamavu WhatsApp Number 7001035870 Meeting With Bangalore Escorts
Call Girls Horamavu WhatsApp Number 7001035870 Meeting With Bangalore EscortsCall Girls Horamavu WhatsApp Number 7001035870 Meeting With Bangalore Escorts
Call Girls Horamavu WhatsApp Number 7001035870 Meeting With Bangalore Escortsvidya singh
 
Call Girls Ludhiana Just Call 9907093804 Top Class Call Girl Service Available
Call Girls Ludhiana Just Call 9907093804 Top Class Call Girl Service AvailableCall Girls Ludhiana Just Call 9907093804 Top Class Call Girl Service Available
Call Girls Ludhiana Just Call 9907093804 Top Class Call Girl Service AvailableDipal Arora
 
Premium Call Girls Cottonpet Whatsapp 7001035870 Independent Escort Service
Premium Call Girls Cottonpet Whatsapp 7001035870 Independent Escort ServicePremium Call Girls Cottonpet Whatsapp 7001035870 Independent Escort Service
Premium Call Girls Cottonpet Whatsapp 7001035870 Independent Escort Servicevidya singh
 
Night 7k to 12k Chennai City Center Call Girls 👉👉 7427069034⭐⭐ 100% Genuine E...
Night 7k to 12k Chennai City Center Call Girls 👉👉 7427069034⭐⭐ 100% Genuine E...Night 7k to 12k Chennai City Center Call Girls 👉👉 7427069034⭐⭐ 100% Genuine E...
Night 7k to 12k Chennai City Center Call Girls 👉👉 7427069034⭐⭐ 100% Genuine E...hotbabesbook
 
Call Girls Bangalore Just Call 9907093804 Top Class Call Girl Service Available
Call Girls Bangalore Just Call 9907093804 Top Class Call Girl Service AvailableCall Girls Bangalore Just Call 9907093804 Top Class Call Girl Service Available
Call Girls Bangalore Just Call 9907093804 Top Class Call Girl Service AvailableDipal Arora
 
💎VVIP Kolkata Call Girls Parganas🩱7001035870🩱Independent Girl ( Ac Rooms Avai...
💎VVIP Kolkata Call Girls Parganas🩱7001035870🩱Independent Girl ( Ac Rooms Avai...💎VVIP Kolkata Call Girls Parganas🩱7001035870🩱Independent Girl ( Ac Rooms Avai...
💎VVIP Kolkata Call Girls Parganas🩱7001035870🩱Independent Girl ( Ac Rooms Avai...Taniya Sharma
 
Call Girls Service Surat Samaira ❤️🍑 8250192130 👄 Independent Escort Service ...
Call Girls Service Surat Samaira ❤️🍑 8250192130 👄 Independent Escort Service ...Call Girls Service Surat Samaira ❤️🍑 8250192130 👄 Independent Escort Service ...
Call Girls Service Surat Samaira ❤️🍑 8250192130 👄 Independent Escort Service ...CALL GIRLS
 
The Most Attractive Hyderabad Call Girls Kothapet 𖠋 6297143586 𖠋 Will You Mis...
The Most Attractive Hyderabad Call Girls Kothapet 𖠋 6297143586 𖠋 Will You Mis...The Most Attractive Hyderabad Call Girls Kothapet 𖠋 6297143586 𖠋 Will You Mis...
The Most Attractive Hyderabad Call Girls Kothapet 𖠋 6297143586 𖠋 Will You Mis...chandars293
 
Vip Call Girls Anna Salai Chennai 👉 8250192130 ❣️💯 Top Class Girls Available
Vip Call Girls Anna Salai Chennai 👉 8250192130 ❣️💯 Top Class Girls AvailableVip Call Girls Anna Salai Chennai 👉 8250192130 ❣️💯 Top Class Girls Available
Vip Call Girls Anna Salai Chennai 👉 8250192130 ❣️💯 Top Class Girls AvailableNehru place Escorts
 
Russian Call Girls in Jaipur Riya WhatsApp ❤8445551418 VIP Call Girls Jaipur
Russian Call Girls in Jaipur Riya WhatsApp ❤8445551418 VIP Call Girls JaipurRussian Call Girls in Jaipur Riya WhatsApp ❤8445551418 VIP Call Girls Jaipur
Russian Call Girls in Jaipur Riya WhatsApp ❤8445551418 VIP Call Girls Jaipurparulsinha
 
Call Girls Coimbatore Just Call 9907093804 Top Class Call Girl Service Available
Call Girls Coimbatore Just Call 9907093804 Top Class Call Girl Service AvailableCall Girls Coimbatore Just Call 9907093804 Top Class Call Girl Service Available
Call Girls Coimbatore Just Call 9907093804 Top Class Call Girl Service AvailableDipal Arora
 
♛VVIP Hyderabad Call Girls Chintalkunta🖕7001035870🖕Riya Kappor Top Call Girl ...
♛VVIP Hyderabad Call Girls Chintalkunta🖕7001035870🖕Riya Kappor Top Call Girl ...♛VVIP Hyderabad Call Girls Chintalkunta🖕7001035870🖕Riya Kappor Top Call Girl ...
♛VVIP Hyderabad Call Girls Chintalkunta🖕7001035870🖕Riya Kappor Top Call Girl ...astropune
 

Recently uploaded (20)

Call Girls Service Jaipur Grishma WhatsApp ❤8445551418 VIP Call Girls Jaipur
Call Girls Service Jaipur Grishma WhatsApp ❤8445551418 VIP Call Girls JaipurCall Girls Service Jaipur Grishma WhatsApp ❤8445551418 VIP Call Girls Jaipur
Call Girls Service Jaipur Grishma WhatsApp ❤8445551418 VIP Call Girls Jaipur
 
Top Rated Bangalore Call Girls Richmond Circle ⟟ 8250192130 ⟟ Call Me For Gen...
Top Rated Bangalore Call Girls Richmond Circle ⟟ 8250192130 ⟟ Call Me For Gen...Top Rated Bangalore Call Girls Richmond Circle ⟟ 8250192130 ⟟ Call Me For Gen...
Top Rated Bangalore Call Girls Richmond Circle ⟟ 8250192130 ⟟ Call Me For Gen...
 
Call Girls Kochi Just Call 9907093804 Top Class Call Girl Service Available
Call Girls Kochi Just Call 9907093804 Top Class Call Girl Service AvailableCall Girls Kochi Just Call 9907093804 Top Class Call Girl Service Available
Call Girls Kochi Just Call 9907093804 Top Class Call Girl Service Available
 
Call Girls Aurangabad Just Call 9907093804 Top Class Call Girl Service Available
Call Girls Aurangabad Just Call 9907093804 Top Class Call Girl Service AvailableCall Girls Aurangabad Just Call 9907093804 Top Class Call Girl Service Available
Call Girls Aurangabad Just Call 9907093804 Top Class Call Girl Service Available
 
Chandrapur Call girls 8617370543 Provides all area service COD available
Chandrapur Call girls 8617370543 Provides all area service COD availableChandrapur Call girls 8617370543 Provides all area service COD available
Chandrapur Call girls 8617370543 Provides all area service COD available
 
Call Girls Bhubaneswar Just Call 9907093804 Top Class Call Girl Service Avail...
Call Girls Bhubaneswar Just Call 9907093804 Top Class Call Girl Service Avail...Call Girls Bhubaneswar Just Call 9907093804 Top Class Call Girl Service Avail...
Call Girls Bhubaneswar Just Call 9907093804 Top Class Call Girl Service Avail...
 
(Rocky) Jaipur Call Girl - 09521753030 Escorts Service 50% Off with Cash ON D...
(Rocky) Jaipur Call Girl - 09521753030 Escorts Service 50% Off with Cash ON D...(Rocky) Jaipur Call Girl - 09521753030 Escorts Service 50% Off with Cash ON D...
(Rocky) Jaipur Call Girl - 09521753030 Escorts Service 50% Off with Cash ON D...
 
VIP Service Call Girls Sindhi Colony 📳 7877925207 For 18+ VIP Call Girl At Th...
VIP Service Call Girls Sindhi Colony 📳 7877925207 For 18+ VIP Call Girl At Th...VIP Service Call Girls Sindhi Colony 📳 7877925207 For 18+ VIP Call Girl At Th...
VIP Service Call Girls Sindhi Colony 📳 7877925207 For 18+ VIP Call Girl At Th...
 
Call Girls Horamavu WhatsApp Number 7001035870 Meeting With Bangalore Escorts
Call Girls Horamavu WhatsApp Number 7001035870 Meeting With Bangalore EscortsCall Girls Horamavu WhatsApp Number 7001035870 Meeting With Bangalore Escorts
Call Girls Horamavu WhatsApp Number 7001035870 Meeting With Bangalore Escorts
 
Call Girls Ludhiana Just Call 9907093804 Top Class Call Girl Service Available
Call Girls Ludhiana Just Call 9907093804 Top Class Call Girl Service AvailableCall Girls Ludhiana Just Call 9907093804 Top Class Call Girl Service Available
Call Girls Ludhiana Just Call 9907093804 Top Class Call Girl Service Available
 
Premium Call Girls Cottonpet Whatsapp 7001035870 Independent Escort Service
Premium Call Girls Cottonpet Whatsapp 7001035870 Independent Escort ServicePremium Call Girls Cottonpet Whatsapp 7001035870 Independent Escort Service
Premium Call Girls Cottonpet Whatsapp 7001035870 Independent Escort Service
 
Night 7k to 12k Chennai City Center Call Girls 👉👉 7427069034⭐⭐ 100% Genuine E...
Night 7k to 12k Chennai City Center Call Girls 👉👉 7427069034⭐⭐ 100% Genuine E...Night 7k to 12k Chennai City Center Call Girls 👉👉 7427069034⭐⭐ 100% Genuine E...
Night 7k to 12k Chennai City Center Call Girls 👉👉 7427069034⭐⭐ 100% Genuine E...
 
Call Girls Bangalore Just Call 9907093804 Top Class Call Girl Service Available
Call Girls Bangalore Just Call 9907093804 Top Class Call Girl Service AvailableCall Girls Bangalore Just Call 9907093804 Top Class Call Girl Service Available
Call Girls Bangalore Just Call 9907093804 Top Class Call Girl Service Available
 
💎VVIP Kolkata Call Girls Parganas🩱7001035870🩱Independent Girl ( Ac Rooms Avai...
💎VVIP Kolkata Call Girls Parganas🩱7001035870🩱Independent Girl ( Ac Rooms Avai...💎VVIP Kolkata Call Girls Parganas🩱7001035870🩱Independent Girl ( Ac Rooms Avai...
💎VVIP Kolkata Call Girls Parganas🩱7001035870🩱Independent Girl ( Ac Rooms Avai...
 
Call Girls Service Surat Samaira ❤️🍑 8250192130 👄 Independent Escort Service ...
Call Girls Service Surat Samaira ❤️🍑 8250192130 👄 Independent Escort Service ...Call Girls Service Surat Samaira ❤️🍑 8250192130 👄 Independent Escort Service ...
Call Girls Service Surat Samaira ❤️🍑 8250192130 👄 Independent Escort Service ...
 
The Most Attractive Hyderabad Call Girls Kothapet 𖠋 6297143586 𖠋 Will You Mis...
The Most Attractive Hyderabad Call Girls Kothapet 𖠋 6297143586 𖠋 Will You Mis...The Most Attractive Hyderabad Call Girls Kothapet 𖠋 6297143586 𖠋 Will You Mis...
The Most Attractive Hyderabad Call Girls Kothapet 𖠋 6297143586 𖠋 Will You Mis...
 
Vip Call Girls Anna Salai Chennai 👉 8250192130 ❣️💯 Top Class Girls Available
Vip Call Girls Anna Salai Chennai 👉 8250192130 ❣️💯 Top Class Girls AvailableVip Call Girls Anna Salai Chennai 👉 8250192130 ❣️💯 Top Class Girls Available
Vip Call Girls Anna Salai Chennai 👉 8250192130 ❣️💯 Top Class Girls Available
 
Russian Call Girls in Jaipur Riya WhatsApp ❤8445551418 VIP Call Girls Jaipur
Russian Call Girls in Jaipur Riya WhatsApp ❤8445551418 VIP Call Girls JaipurRussian Call Girls in Jaipur Riya WhatsApp ❤8445551418 VIP Call Girls Jaipur
Russian Call Girls in Jaipur Riya WhatsApp ❤8445551418 VIP Call Girls Jaipur
 
Call Girls Coimbatore Just Call 9907093804 Top Class Call Girl Service Available
Call Girls Coimbatore Just Call 9907093804 Top Class Call Girl Service AvailableCall Girls Coimbatore Just Call 9907093804 Top Class Call Girl Service Available
Call Girls Coimbatore Just Call 9907093804 Top Class Call Girl Service Available
 
♛VVIP Hyderabad Call Girls Chintalkunta🖕7001035870🖕Riya Kappor Top Call Girl ...
♛VVIP Hyderabad Call Girls Chintalkunta🖕7001035870🖕Riya Kappor Top Call Girl ...♛VVIP Hyderabad Call Girls Chintalkunta🖕7001035870🖕Riya Kappor Top Call Girl ...
♛VVIP Hyderabad Call Girls Chintalkunta🖕7001035870🖕Riya Kappor Top Call Girl ...
 

Ultrasound nerve segmentation, kaggle review

  • 1. Kaggle ultrasound nerve segmentation Tyantov Eduard
  • 2. #1 Description Thu 19 May 2016 – Thu 18 Aug 2016
  • 3. #2 Data: to find Brachial Plexus (BP) – 420x580 resolution – 5635 train images with masks, 5508 test; – ~120 images per 47 patients – 47% of the images don’t have a mask; – result in RLE encoding
  • 4. #3 Data: mistakes in the ground truth – 45 known errors of near duplicate images – metric is sensitive to nerve presence mistakes
  • 5. #4 Evaluation Peculiarities – (!) Mask presence mistake leads to zero score – Needs smoothing in the denominator Loss functions – 1 – dice – -dice – weighted cross entropy (2 classes, per pixel prediction) Mean mask
  • 6. #5 Baselines Score Description Framework Author 0.51 Empty submission - - 0.00 Top left pixel - - 0.57 U-Net, in the beginning of the competition Keras code Marko Jocic, kaggler 0.62 U-Net, almost at the end Torch code Qure.ai (host)
  • 7. #6 What is U-Net Overview – (May 2015 article) “U-Net: Convolutional Networks for Biomedical Image Segmentation” – Winner of “Grand Challenge for Computer-Automated Detection of Caries in Bitewing Radiography at ISBI 2015” – Encoder-decoder architecture with skip connection on the same level – Fully convolutional, Drop-out in the middle – Augmentation: “Smooth deformations using random displacement vectors on a coarse 3 by 3 grid. The displacements are sampled from a Gaussian distribution with 10 pixels standard deviation.”
  • 8. #7 Another aproach, FCN Overview – (20 May 2016, article) “Fully Convolutional Networks for Semantic Segmentation” – VGG-18 – Segmentation prediction on different layers of the net, +upsampling – Average predictions
  • 9. #8 Starting point, Marko Jocic’s solution Overview – Classic U-Net: VGG-like – Very simple Keras code – Image resize to 64x80, bicubic interpolation – Loss= - Dice coefficient, per batch averaging, smooth=1 – Training on whole dataset, no validation – RLE-encoding function – Adam optimizer Training – 20 epochs, ~30 seconds on Titan X, memory footprint 800mb – Overfits, 0.68 on training -> 0.57 on LB
  • 10. #9 Aspects of the solution: basics Overfitting basics (+2%) – Split train/valid, 20% and early stopping patience=5 epochs • used random split instead of more convenient by patient (due to a subtle bug) – Dropout after each conv layer General enhancements – Resolution 64x80 -> 80x112 (+1%) – ELU instead of ReLU -> faster convergence
  • 11. #10 Aspects of the solution: augmentation Augmentation* – flip x, y – random rotate (5) – random zoom (0.9, 1.1) – random channel shift (5.0) *all transformations should be done with a mask too All transformations can be done on the fly with a generator (randomly applied), but didn’t improve results. Elastic transform: convolve with a Gaussian on random displacement fields Result: no added effect
  • 12. #11 Aspects of the solution: blocks Modifications of U-Net – 2 3x3 convolution -> inception_v3 block – BNA after each convolution – BNA + activation after summation – nxn -> 1xn + nx1 Results: – lesser parameters (1M) – faster convergence – LB: +2% v3 block v3 + splitted
  • 13. #12 Aspects of the solution: 2nd head, postfilter 2nd head – mask presence branch in the middle of NN (after decoder part) • Conv 1x1, sigmoid • FC=1, sigmoid – leads to better convergence Post filter – presence prob < 0.5 or sum(pixels) < 3000 -> empty mask (+4.5%) – in the end: combining p_nerve = (p_score + p_segment)/2 -> +0.5%
  • 14. #13 Aspects of the solution: other Modifications – Skip connection with Residual blocks (+1%) – Max pool -> Conv 3x3 with stride=2 (+BNA) – Ensemble (+1%) • k-fold 5,6,7,8, average – Prediction on augmented versions of test images (averaging) Final result: – single model 0.694 score – ensemble 0.70399 (hour before the competition’s end) – last submission has been human verified ;) but no help code: https://github.com/EdwardTyantov/ultrasound-nerve-segmentation
  • 15. #14 Leaderboard – 31-th – see Private LB huge shake-up
  • 16. #15 What didn’t help – Inception Resnet v4 – sequential training of decoder, encoder parts – more or lesser layers/blocks/n_filters – pixel clustering – higher or lower resolution – dropout, different probs – Torch version – deconv layers instead of upsampling – weight decay for layers – FCN – Deepmask architecture
  • 17. #16 Technical – Ubuntu 14 or 16, Cuda 8, cudnn 5, keras last, torch last – batch_size=64, 128 (depends on GPU memory) – Single model, 2-3 hours on Titan/1080 – Ensemble – 24 hours
  • 18. – train dataset: error re-labeling or zero-outing – FCN with several heads in different resolutions (regularize) – post process: mask to elipse, no holes – separate training: mask/no mask – crop images, super-resolution – models on different resolution – higher resolution – loss – smart post-processing • which obv. led to overfitting on Public Score – replication padding instead of zero-padding #17 Other competitors
  • 19. #18 Deepmask (FB) – no low-level features – CNN with two heads: mask and score – training set: patch, mask, y_k - objected in centered and fully contained • mask pixel=1 if it is part of object in the center – VGG, 8-layers, can be trained – Training • joint learning score * 1/32 • first branch only positives • augmentation: shift 16pix, scale little, horiz. flip – Evaluation • full image: 16 pixel stride
  • 21. #20 Sharpmask (refined deepmask) – Architecture • trunk: resnet-50 pretrained • reflect-pad instead of zero • for DeepMask: headC – Training • Deepmask (score + coarse mask) • then freeze and learn refinements • Why: faster convergenc, deepmask or sharpmask result, minimal finetuning – Inference • only top-n locations refined