Final project in KAIST AI class

Overview

mmodal_mixer

MLP-Mixer based Multi-modal image-text retrieval

Image:

Original image is cropped with 16 x 16 patch size without overlap. Then, it is reshaped to (batch, (hxw), (patch x patch x channel)).

Text:

Also, original text is tokenized and embedded with BERT-based approach (BERT-base-uncased).

Data processing:

When we train our model, we randomly samples(50 %) reports to make the matched- and un-matched image-text set. Basically, matched and un-matched set is classified with label information using chexpert labeler, we consider unmatched set when randomly sampled report is not exactly same with original one.

Mixer based approach is trained efficiently with xxxx throuput with xxx accuracy.

Exp settings.

batch: 256 batch epoch: 50 epoch

Chest X-ray Image-reports retrieval

Model spec: patch size:16, embedding dim: 768

Input spec: img size: 224x224x3 -> pathch size: (224/16) x (224/16) text max len: 128 legth

input embedding: cls, txt, sep, img output: matched or unmatched

Results

Max memory: 15.903G

Model forward time: Model backward time:

Owner
SuperSuperMoon
PhD student at Graduate School of AI, KAIST. Previously AI research engineer at Samsung Medical Center. Medical AI. Computer Vision & NLP.
SuperSuperMoon
Kellogg bad | Union good | Support strike funds

KelloggBot Credit to SeanDaBlack for the basis of the script. req.py is selenium python bot. sc.js is a the base of the ios shortcut [COMING SOON] Set

407 Nov 17, 2022
A plugin for poetry that allows you to execute scripts defined in your pyproject.toml, just like you can in npm or pipenv

poetry-exec-plugin A plugin for poetry that allows you to execute scripts defined in your pyproject.toml, just like you can in npm or pipenv Installat

38 Jan 06, 2023
Automatically unpin old messages so you can always pin more!

PinRotate Automatically unpin old messages so you can always pin more! Installation You will need to install poetry to run this bot locally for develo

3 Sep 18, 2022
Predict if a fuse is usable on an appliance depending on the fuse rating

fuse-feasibility-analysis Predict if a fuse is usable on an appliance depending on the fuse rating , Power rating and resistance in the appliance

Sebastian Muchui 4 Jul 21, 2022
Dapp / Forge traces enhancer

traces-explorer Dapp / Forge traces enhancer Usage traces.py and pattern_* files should be in the same directory make test traces.txt py traces.

1 Feb 02, 2022
A program that takes Python classes and turns them into CSS classes.

PyCSS What is it? PyCSS is a micro-framework to speed up the process of writing bulk CSS classes. How does it do it? With Python!!! First download the

T.R Batt 0 Aug 03, 2021
⏰ Shutdown Timer is an application that you can shutdown, restart, logoff, and hibernate your computer with a timer.

Shutdown Timer is a an application that you can shutdown, restart, logoff, and hibernate your computer with a timer. After choosing an action from the

Mehmet Güdük 5 Jun 27, 2022
An addin for Autodesk Fusion 360 that lets you view your design in a Looking Glass Portrait 3D display

An addin for Autodesk Fusion 360 that lets you view your design in a Looking Glass Portrait 3D display

Brian Peiris 12 Nov 02, 2022
Automatically find solutions when your Python code encounters an issue.

What The Python?! Helping you find answers to the errors Python spits out. Installation You can find the source code on GitHub at: https://github.com/

What The Python?! 139 Dec 14, 2022
An example project that shows how to check if a certain macro is active in a file.

PlatformIO Check Compiler Flags Example Description Demonstrates the usage of an extra script and a special compilter invocation to get the active mac

Maximilian Gerhardt 1 Oct 28, 2021
Kunai Shitty Raider Leaked LMFAO

Kunai-Raider-Leaked Kunai Shitty Raider Leaked LMFA

5 Nov 24, 2021
使用clash核心,对服务器进行Netflix解锁批量测试。

注意事项 测速及解锁测试仅供参考,不代表实际使用情况,由于网络情况变化、Netflix封锁及ip更换,测速具有时效性 本项目使用 Python 编写,使用前请完成环境安装 首次运行前请安装pip及相关依赖,也可使用 pip install -r requirements.txt 命令自行安装 Net

11 Dec 07, 2022
Tools for dos (denial-of-service) website / web server

DoS Attack Tools Tools for dos (denial-of-service) website / web server di buat olah NurvySec How to install on debian / ubuntu $ apt update $ apt ins

nurvy 1 Feb 10, 2022
Manipulation OpenAI Gym environments to simulate robots at the STARS lab

liegroups Python implementation of SO2, SE2, SO3, and SE3 matrix Lie groups using numpy or PyTorch. [Documentation] Installation To install, cd into t

STARS Laboratory 259 Dec 11, 2022
通过简单的卷积神经网络直接预测出验证码图片中滑块的位置

使用说明 1. 在本地测试 运行python3 prdict_one.py即可,默认需要预测的图片路径位于testImg文件夹下的test1.png 运行python3 predict_folder.py预测testImg下的所有图片 2. 部署到服务器 运行python3 run_a_server

12 Mar 08, 2022
Its a simple and fun to use application. You can make your own quizes and send the lik of the quiz to your friends.

Quiz Application Its a simple and fun to use application. You can make your own quizes and send the lik of the quiz to your friends. When they would a

Atharva Parkhe 1 Feb 23, 2022
CoreSE - basic of social Engineering tool

Core Social Engineering basic of social Engineering tool. just for fun :) About First of all, I must say that I wrote such a project because of my int

Hamed Mohammadvand 7 Jun 10, 2022
A water drinking notification every hour to keep you healthy while coding :)

Water_Notification A water drinking notification every hour to keep you healthy while coding. 💧 💧 Stay Hydrated Stay Healthy 💧 💧 Authors @CrazyCat

Arghya Banerjee 1 Dec 22, 2021
DRF magic links

drf-magic-links Installation pip install drf-magic-links Add URL patterns # urls.py

Dmitry Kalinin 1 Nov 07, 2021
Simple plug-and-play installer for users who want to LineageOS from stock firmware, or from another custom ROM.

LineageOS for the Teracube 2e Simple plug-and-play installer for users who want to LineageOS from stock firmware, or from another custom ROM. Dependen

Gagan Malvi 5 Mar 31, 2022