1 Star 5 Fork 1

cubone/learnopencv

加入 Gitee
与超过 1200万 开发者一起发现、参与优秀开源项目,私有仓库也完全免费 :)
免费加入
文件
AgeGender
AlphaBlending
AugmentedRealityWithArucoMarkers
Autoencoder-in-TensorFlow
Background-Subtraction
Bag-Of-Tricks-For-Image-Classification
BatchNormalization
BlobDetector
CI
CameraCalibration
CenterofBlob
CharClassification
Classification-with-localization-convert-any-keras-classi...
ColorSpaces
Colorization
Colormap
Contour-Detection-using-OpenCV
ConvexHull
Creating-a-Virtual-Pen-and-Eraser
DNN-OpenCV-Classification-Android
DNN-OpenCV-Classification-with-Java
Deep-Learning-with-OpenCV-DNN-Module
Delaunay
DenoisingAutoencoder
Depth-Perception-Using-Stereo-Camera
Efficient-image-loading
EfficientNet
EigenFace
EpipolarGeometryAndStereoVision
ExposureFusion
FBAMatting
FPS
Face-Recognition-with-ArcFace
FaceAverage
FaceDetectionComparison
FaceMaskOverlay
FaceMorph
FaceSwap
FacialAttractiveness
FacialLandmarkDetection
Federated-Learning-Intro
FunnyMirrors
GOTURN
Getting-Started-OpenCV-CUDA-Module
Graph-Convolutional-Networks-Model-Relations-In-Data
HandPose
Hangman
HeadPose
Holes
Homography
Hough-Transform
HuMoments
Image-Classification-in-PyTorch
Image-Inpainting
Image-classification-pre-trained-models
ImageAlignment-FeatureBased
ImageAlignment
ImageMetrics
Improving-Illumination-in-Night-Time-Images
Inference-for-PyTorch-Models/ONNX-Caffe2
Install-OpenCV-Windows-exe
InstallScripts
Intro-to-Generative-Adversarial-Network
InvisibilityCloak
Keras-Fine-Tuning
Keras-ImageNet-Models
Keras-Transfer-Learning
KerasCNN-CIFAR
KerasMLP-MNIST
MRNet-Single-Model
MRnet-MultiTask-Approach
content
dataset
models
src
utils
.gitignore
LICENSE
README.md
config.py
train.py
Mask-RCNN
MultiObjectTracker
Non-Maximum-Suppression
NonPhotorealisticRendering
OCR
ObjectDetection-YOLO
OpenCV-dnn-gpu-support-Linux
OpenCV-dnn-gpu-support-Windows
OpenPose-Multi-Person
OpenPose
OpenVINO-OpenCV
Optical-Flow-Estimation-using-Deep-Learning-RAFT
Optical-Flow-in-OpenCV
Photoshop-Filters-in-OpenCV
Playing-Chrome-TRex-Game-with-Facial-Gestures
Playing-rock-paper-scissors-with-AI
PyTorch-Fully-Convolutional-Image-Classification
PyTorch-Keypoint-RCNN
PyTorch-Mask-RCNN
PyTorch-Multi-Label-Image-Classification-Image-Tagging
PyTorch-Multi-Label-Image-Classification
PyTorch-ONNX-TensorRT-CPP
PyTorch-ONNX-TensorRT
PyTorch-Receptive-Field-With-Backprop
PyTorch-Segmentation-torchvision
PyTorch-Vision-Experiment-Logging
PyTorch-faster-RCNN
PyTorch-for-Beginners
PyTorch-to-CoreML-model-conversion
PyTorch-to-TensorFlow-Model-Conversion
Pytorch-Lightning
Pytorch-Net-Surgery
QRCode-OpenCV
Quora-Dataset-Duplicate-Search
ReconstructFaceUsingEigenFaces
RedEyeRemover
RotationMatrixToEulerAngles
SVM-using-Python
SeamlessCloning
SelectiveSearch
SnakeGame
Super-Resolution-in-OpenCV
TSNE
TensorBoard-With-Pytorch-Lightning
TensorFlow-Fully-Convolutional-Image-Classification
TensorFlow-Receptive-Field-With-Backprop
Tetris
TextDetectionEAST
Threshold
Training_a_custom_hand_detector_with_dlib
UI-cvui
UnderstandingLensDistortion
Universal-Sentence-Encoder
Variational-Autoencoder-TensorFlow
VideoBackgroundEstimation
VideoReadWriteDisplay
VideoStabilization
WarpTriangle
XeusCling
YOLOv3-Training-Snowman-Detector
app-seperation-semseg
barcode-QRcodeScanner
digits-classification
docs
downloadOpenImages
forEach
hdr
heatmap
imread
keras-linear-regression
ninjaEyeDetector
otsu-method
pymodule
qt-test
stereo-camera
tracking
video-classification-and-human-activity-recognition
.gitignore
README.md
该仓库未声明开源许可证文件(LICENSE),使用请关注具体项目描述及其代码上游依赖。
克隆/下载
贡献代码
同步代码
取消
提示: 由于 Git 不支持空文件夾,创建文件夹后会生成空的 .keep 文件
Loading...
README

Stanford MRnet Challenge

This repo contains code for the MRNet Challenge (A Multi-Task Approach)

For more details refer to https://stanfordmlgroup.github.io/competitions/mrnet/

Install dependencies

  1. pip install git+https://github.com/ncullen93/torchsample
  2. pip install nibabel
  3. pip install sklearn
  4. pip install pandas

Install other dependencies as per requirement

Instructions to run the training

  1. Clone the repository.

  2. Download the dataset (~5.7 GB), and put train and valid folders along with all the the .csv files inside images folder at root directory.

  images/
      train/
          axial/
          sagittal/
          coronal/
      val/
          axial/
          sagittal/
          coronal/
      train-abnormal.csv
      train-acl.csv
      train-meniscus.csv
      valid-abnormal.csv
      valid-acl.csv
      valid-meniscus.csv
  1. Make a new folder called weights at root directory, and inside the weights folder create three more folders namely acl, abnormal and meniscus.

  2. All the hyperparameters are defined in config.py file. Feel free to play around those.

  3. Now finally run the training using python train.py. All the logs for tensorboard will be stored in the runs directory at the root of the project.

Understanding the Dataset

The dataset contains MRIs of different people. Each MRI consists of multiple images. Each MRI has data in 3 perpendicular planes. And each plane as variable number of slices.

Each slice is an 256x256 image

For example:

For MRI 1 we will have 3 planes:

Plane 1- with 35 slices

Plane 2- with 34 slices

Place 3 with 35 slices

Each MRI has to be classisifed against 3 diseases

Major challenge with while selecting the model structure was the inconsistency in the data. Although the image size remains constant , the number of slices per plane are variable within a single MRI and varies across all MRIs.

Model Specifications

In the last attempt to MRNet challenge, we used 3 different models for each disease, but instead we can leverage the information that the model learns for each of the disease and make inferencing for other disease better.

We used Hard Parameter sharing in this approach.

We will be using 3 Alexnet pretrained as 3 feature extractors for each of the plane. We then combine these feature extractor layers as an input to a global fully connected layer for the final classification.

Contributors

-- Neelabh Madan

-- Jatin Prakash

马建仓 AI 助手
尝试更多
代码解读
代码找茬
代码优化
1
https://gitee.com/cubone/learnopencv.git
git@gitee.com:cubone/learnopencv.git
cubone
learnopencv
learnopencv
master

搜索帮助