Effect of Deep Transfer and Multi task Learning on Sperm Abnormality Detection

Last update: Sep 05, 2022

Overview

Effect of Deep Transfer and Multi task Learning on Sperm Abnormality Detection

Introduction

This repository includes codes and models of "Effect of Deep Transfer and Multi task Learning on Sperm Abnormality Detection" paper. link: https://doi.org/10.1016/j.compbiomed.2020.104121

Dataset

First you should download the MHSMA dataset using:

git clone https://github.com/soroushj/mhsma-dataset.git

Usage

First of all,the configuration file should be setted.So open dmtl.txt or dtl.txt and set the setting you want.This files contains paramaters of the model that you are going to train.

dtl.txt have only one line and contains paramaters to train a DTL model.
dmtl.txt contains two lines:paramaters of stage 1 are kept in the first line of the file and paramaters of stage 2 are kept in the second line of the file.
Some paramaters have an aray of three values that they keep the value of three labels.To set them,consider this sequence:[Acrosome,Vacoule,Head].
To train a DTL model,use the following commands and arguments:

python train.py -t dtl [-e epchos] [-label label]  [-model model] [-w file]

Argumetns:

Argument	Description
-t	type of network(dtl or dmtl)
-e	number of epochs
-label	label(a,v or h)
-model	pre-trained model
-w	name of best weihgt file
--phase	You can use it to choose stage in DMTL(1 or 2)
--second_model	The base model for second stage of DMTL

1.Train

To choose a pre-trained model, you can use one of the following models:

model argument	Description
vgg_19	VGG 19
vgg_16	VGG 16
resnet_50	Resnet 50
resnet_101	Resnet 101
resnet_502	Resnet 502

To train a DMTL model,use the following commands and arguments:

python train.py -t dmtl [--phase phase] [-e epchos] [-label label] [-model model] [-w file]

Also you can use your own pre-trained model by using address of your model instead of the paramaters been told in the table above.

Example:
python train.py -t dmtl --phase 1 -e 100 -label a -model C:\model.h5 -w w.h5

2.K Fold

To perform K Fold on a model,use "-k_fold True" argument.

python train.py -k_fold True [-t type] [-e epchos] [-label label] [-model model] [-w file]

3.Threshold Search

To find a good threshold for your model,use the following code:

python threshold.py [-t type] [-addr model address] [-l label]

Models

The CNN models that were introduced and evaluated in our research paper can be found in the v1.0 release of this repository.

You might also like...

Face Detection and Alignment using Multi-task Cascaded Convolutional Networks (MTCNN)

Face-Detection-with-MTCNN Face detection is a computer vision problem that involves finding faces in photos. It is a trivial problem for humans to sol

3 Oct 7, 2022

Multi-task yolov5 with detection and segmentation based on yolov5

YOLOv5DS Multi-task yolov5 with detection and segmentation based on yolov5(branch v6.0) decoupled head anchor free segmentation head README中文 Ablation

150 Dec 30, 2022

Code for the ICML 2021 paper "Bridging Multi-Task Learning and Meta-Learning: Towards Efficient Training and Effective Adaptation", Haoxiang Wang, Han Zhao, Bo Li.

Bridging Multi-Task Learning and Meta-Learning Code for the ICML 2021 paper "Bridging Multi-Task Learning and Meta-Learning: Towards Efficient Trainin

57 Dec 15, 2022

A novel Engagement Detection with Multi-Task Training (ED-MTT) system

Comments

a possible typo(bug)

Very interesting idea and complements!

In LoadData.py, starting from line 150, ` if phase == 'search':

    return {
            "x_train": x_train_128,
            "y_train": y_train,
            "x_train_128": x_train_128,
            'x_val_128': x_valid_128,
            "x_val": x_valid_128,
            "y_val": y_valid,
            "x_test": x_test_128,
            "y_test": y_test
            }`

here, I think that the first key-value pair should probably be "x_train": x_train instead of "x_train": x_train_128, which causes an error of shape mismatch during fit.

opened by captainst 0

Effect of Deep Transfer and Multi task Learning on Sperm Abnormality Detection

Related tags

Overview

Effect of Deep Transfer and Multi task Learning on Sperm Abnormality Detection

Introduction

Dataset

Usage

1.Train

2.K Fold

3.Threshold Search

Models

You might also like...

Face Detection and Alignment using Multi-task Cascaded Convolutional Networks (MTCNN)

Multi-task yolov5 with detection and segmentation based on yolov5

Code for the ICML 2021 paper "Bridging Multi-Task Learning and Meta-Learning: Towards Efficient Training and Effective Adaptation", Haoxiang Wang, Han Zhao, Bo Li.

A novel Engagement Detection with Multi-Task Training (ED-MTT) system

Self-training for Few-shot Transfer Across Extreme Task Differences

Effect of Different Encodings and Distance Functions on Quantum Instance-based Classifiers

Efficient neural networks for analog audio effect modeling

[CVPR 2021] Counterfactual VQA: A Cause-Effect Look at Language Bias

Algebraic effect handlers in Python

Comments

a possible typo(bug)

Releases(v1.0)

v1.0(Dec 29, 2020)

Owner

Amir Abbasi

pytorch implementation of ABC : Auxiliary Balanced Classifier for Class-imbalanced Semi-supervised Learning

mlpack: a scalable C++ machine learning library --

This repository contains notebook implementations of the following Neural Process variants: Conditional Neural Processes (CNPs), Neural Processes (NPs), Attentive Neural Processes (ANPs).

A high-level Python library for Quantum Natural Language Processing

Code of paper "CDFI: Compression-Driven Network Design for Frame Interpolation", CVPR 2021

Implementation for NeurIPS 2021 Submission: SparseFed

Usable Implementation of "Bootstrap Your Own Latent" self-supervised learning, from Deepmind, in Pytorch

Source code release of the paper: Knowledge-Guided Deep Fractal Neural Networks for Human Pose Estimation.

Codes for NeurIPS 2021 paper "On the Equivalence between Neural Network and Support Vector Machine".

Pytorch Performace Tuning, WandB, AMP, Multi-GPU, TensorRT, Triton

Federated Learning Based on Dynamic Regularization

Fully Convolutional DenseNet (A.K.A 100 layer tiramisu) for semantic segmentation of images implemented in TensorFlow.

The code for our paper "NSP-BERT: A Prompt-based Zero-Shot Learner Through an Original Pre-training Task —— Next Sentence Prediction"

Generate images from texts. In Russian. In PaddlePaddle

On-device speech-to-intent engine powered by deep learning

The code for our paper submitted to RAL/IROS 2022: OverlapTransformer: An Efficient and Rotation-Invariant Transformer Network for LiDAR-Based Place Recognition.

Implements Stacked-RNN in numpy and torch with manual forward and backward functions

Learning to Disambiguate Strongly Interacting Hands via Probabilistic Per-Pixel Part Segmentation [3DV 2021 Oral]

Machine learning for NeuroImaging in Python

[ICCV'21] Learning Conditional Knowledge Distillation for Degraded-Reference Image Quality Assessment