PI-DECODER

Introduction

PI-DECODER, a decoder structure designed for Precise Iris Segmentation and Location. The decoder structure is shown below:

Please check technical paper.pdf in the "reference" subfolder for more details.

How to use?

For african dataset, you can enter the following script on your terminal:

python main.py --mode test --model_path ./models/african_best.pth --test_mode 1 --train_dataset african

Then you have iris mask, pupil mask and outer iris mask that are predicted by the input images. At the same time, the relevant index data will be displayed on your terminal.

(ijcb) PS F:\workspace\code\pytorch\PI-DECODER> python main.py --mode test --model_path ./models/african_best.pth --
test_mode 1 --train_dataset african
Namespace(batch_size=1, beta1=0.9, beta2=0.999, img_size=(640, 640), lr=0.0002, mode='test', model_path='./models/af
rican_best.pth', num_epochs=100, num_workers=2, result_path='./result/', test_mode=1, test_path='./dataset/test/', t
rain_dataset='african', train_path='./dataset/train/', valid_path='./dataset/valid/')
image count in train path :5
image count in valid path :5
image count in test path :40
Using Model: PI-DECODER
0.0688 seconds per image

----------------------------------------------------------------------------------------------------------------
|evaluation     |e1(%)          |e2(%)          |miou(%)        |f1(%)          |miou_back      |f1_back        |
----------------------------------------------------------------------------------------------------------------
|iris seg       |0.384026       |0.192013       |91.175200      |95.350625      |95.386805      |97.574698      |
|iris mask      |0.569627       |0.284813       |93.159855      |96.430411      |96.270919      |98.060105      |
|pupil mask     |0.078793       |0.039396       |93.138878      |96.409347      |96.529547      |98.184718      |
----------------------------------------------------------------------------------------------------------------
|average        |0.344149       |0.172074       |92.491311      |96.063461      |96.062424      |97.939840      |
----------------------------------------------------------------------------------------------------------------

Besides, if you don't have groud-truth files or just want to save the results, use test mode 2.

python main.py --mode test --model_path ./models/african_best.pth --test_mode 2 --train_dataset african

Requirements

The whole experiment was run on the NVIDIA RTX 3060. The following are recommended environment configurations.

matplotlib        3.3.4
numpy             1.19.5
opencv-python     4.5.1.48
pandas            1.1.5
Pillow            8.1.2
pip               21.0.1
pyparsing         2.4.7
python-dateutil   2.8.1
pytz              2021.1
scipy             1.5.4
setuptools        52.0.0.post20210125
six               1.15.0
thop              0.0.31.post2005241907
torch             1.7.0+cu110
torchstat         0.0.7
torchsummary      1.5.1
torchvision       0.8.1+cu110

precise iris segmentation

Related tags

Overview

PI-DECODER

Introduction

How to use?

Requirements

Owner

Simple translation demo showcasing our headliner package.

AMUSE - financial summarization

Coreference resolution for English, German and Polish, optimised for limited training data and easily extensible for further languages

InferSent sentence embeddings

LOT: A Benchmark for Evaluating Chinese Long Text Understanding and Generation

Python library for parsing resumes using natural language processing and machine learning

DANeS is an open-source E-newspaper dataset by collaboration between DATASET JSC (dataset.vn) and AIV Group (aivgroup.vn)

Transformers4Rec is a flexible and efficient library for sequential and session-based recommendation, available for both PyTorch and Tensorflow.

A list of NLP(Natural Language Processing) tutorials

Fuzzy String Matching in Python

The tool to make NLP datasets ready to use

Guide: Finetune GPT2-XL (1.5 Billion Parameters) and GPT-NEO (2.7 B) on a single 16 GB VRAM V100 Google Cloud instance with Huggingface Transformers using DeepSpeed

Code repository of the paper Neural circuit policies enabling auditable autonomy published in Nature Machine Intelligence

Continuously update some NLP practice based on different tasks.

WIT (Wikipedia-based Image Text) Dataset is a large multimodal multilingual dataset comprising 37M+ image-text sets with 11M+ unique images across 100+ languages.

Contains analysis of trends from Fitbit Dataset (source: Kaggle) to see how the trends can be applied to Bellabeat customers and Bellabeat products

Code for "Parallel Instance Query Network for Named Entity Recognition", accepted at ACL 2022.

An easier way to build neural search on the cloud

The implementation of Parameter Differentiation based Multilingual Neural Machine Translation

LSTM based Sentiment Classification using Tensorflow - Amazon Reviews Rating