9th place solution

Overview

AllDataAreExt-Galixir-Kaggle-HPA-2021-Solution

Team Members

  • Qishen Ha is Master of Engineering from the University of Tokyo. Machine Learning Engineer at LINE Corp. Kaggle Competition Grandmaster. Z by HP & NVIDIA Global Data Science Ambassador.

  • Bo Liu is currently a Senior Deep Learning Data Scientist at NVIDIA based in the U.S. and a Kaggle Competition Grandmaster.

  • Fuxu Liu is currently a Algorithm Engineer at ReadSense based in the China. Kaggle Competition Grandmaster. Z by HP & NVIDIA Global Data Science Ambassador.

  • Daishu is currently a Senior Research Scientist at Galixir. Kaggle Competition Grandmaster.

Methods

Overview of Methods

Image-to-cell augmentation module

We used two methods to train and make predictions in our pipeline.

Firstly, we use 512 x 512 image size to train and test. For predicting, we loop n times for each image (n is the number of cells in the image), leaving only one cell in each time and masking out the other cells to get single cell predictions.

The second method is trained with 768 x 786 images with random crop to 512 x 512 then tested almost the same way as our first approach. Specifically, we not only mask out the other cells but reposition of the cells in the left to the center of the image as well.

The two methods share the same training process, in which we incorporate two augmentation approach specifically designed for this task, in addition to regular augmentation methods such as random rotation, flipping, cropping, cutout and brightness adjusting. The first augmentation approach is, with a small probability, multiplying the data of the green channel (protein) by a random number in the range of [0.0,0.1] while setting the label to negative to improve the model's ability to recognize negative samples. The other augmentation approach is, with a small probability, setting the green channel to red (Microtubules) or yellow (Endoplasmicreticulum), multiplying it by a random number in the range of [0.6,1.0] and changing the label to the Microtubules or Endoplasmicreticulum.

pseudo-3D cell augmentation module

We pre-crop all the cells of each image and save them locally. Then during training, for each image we randomly select 16 cells. We then set bs=32, so for each batch we have 32x16=512 cells in total.

We resize each cell to 128x128, so the returned data shape from the dataloader is (32, 16, 4, 128, 128) . Next we reshape it into (512, 4, 128, 128) and then use a very common CNN to forward it, the output shape is (512, 19).

In the prediction phase we use the predicted average of different augmented images of a cell as the predicted value for each cell. But during the training process, we rereshape this (512, 19) prediction back into (32, 16, 19) . Then the loss is calculated for each cell with image-level GT label.

Featurziation with deep neural network

We use multipe CNN variants to train, such as EfficientNet, ResNet, DenseNet.

Classification

We average the different model predictions from different methods.

Tree-Structured Directory

├── input

│   ├──hpa-512: 512-image and 512-cell mask

│   │   ├── test

│   │   ├── test_cell_mask

│   │   ├── train

│   │   └── train_cell_mask

│   ├── hpa-seg : official segmentation models

│   └── hpa-single-cell-image-classification : official data and kaggle_2021.tsv

├── output : logs, models and submission

Code

  • S1_external_data_download.py: download external train data

  • S2_data_process.py: generate 512-image and 512-cell mask

  • S3_train_pipeline1.py: train image-to-cell augmentation module

  • S4.1_crop_cells.py: crop training cells for pseudo-3D cell augmentation module

  • S4.2_train_pipeline2.py: train pseudo-3D cell augmentation module

  • S5_predict.py: generate submission.csv

Owner
daishu
daishu
Lightweight Face Image Quality Assessment

LightQNet This is a demo code of training and testing [LightQNet] using Tensorflow. Uncertainty Losses: IDQ loss PCNet loss Uncertainty Networks: Mobi

Kaen 5 Nov 18, 2022
Repository relating to the CVPR21 paper TimeLens: Event-based Video Frame Interpolation

TimeLens: Event-based Video Frame Interpolation This repository is about the High Speed Event and RGB (HS-ERGB) dataset, used in the 2021 CVPR paper T

Robotics and Perception Group 544 Dec 19, 2022
Code for "Infinitely Deep Bayesian Neural Networks with Stochastic Differential Equations"

Infinitely Deep Bayesian Neural Networks with SDEs This library contains JAX and Pytorch implementations of neural ODEs and Bayesian layers for stocha

Winnie Xu 95 Nov 26, 2021
CodeContests is a competitive programming dataset for machine-learning

CodeContests CodeContests is a competitive programming dataset for machine-learning. This dataset was used when training AlphaCode. It consists of pro

DeepMind 1.6k Jan 08, 2023
Learned image compression

Overview Pytorch code of our recent work A Unified End-to-End Framework for Efficient Deep Image Compression. We first release the code for Variationa

Jiaheng Liu 163 Dec 04, 2022
Collection of generative models in Pytorch version.

pytorch-generative-model-collections Original : [Tensorflow version] Pytorch implementation of various GANs. This repository was re-implemented with r

Hyeonwoo Kang 2.4k Dec 31, 2022
Implementation of "StrengthNet: Deep Learning-based Emotion Strength Assessment for Emotional Speech Synthesis"

StrengthNet Implementation of "StrengthNet: Deep Learning-based Emotion Strength Assessment for Emotional Speech Synthesis" https://arxiv.org/abs/2110

RuiLiu 65 Dec 20, 2022
kapre: Keras Audio Preprocessors

Kapre Keras Audio Preprocessors - compute STFT, ISTFT, Melspectrogram, and others on GPU real-time. Tested on Python 3.6 and 3.7 Why Kapre? vs. Pre-co

Keunwoo Choi 867 Dec 29, 2022
A nutritional label for food for thought.

Lexiscore As a first effort in tackling the theme of information overload in content consumption, I've been working on the lexiscore: a nutritional la

Paul Bricman 34 Nov 08, 2022
StyleGAN2-ada for practice

This version of the newest PyTorch-based StyleGAN2-ada is intended mostly for fellow artists, who rarely look at scientific metrics, but rather need a working creative tool. Tested on Python 3.7 + Py

vadim epstein 170 Nov 16, 2022
Multi-Scale Geometric Consistency Guided Multi-View Stereo

ACMM [News] The code for ACMH is released!!! [News] The code for ACMP is released!!! About ACMM is a multi-scale geometric consistency guided multi-vi

Qingshan Xu 118 Jan 04, 2023
This is the official implementation of 3D-CVF: Generating Joint Camera and LiDAR Features Using Cross-View Spatial Feature Fusion for 3D Object Detection, built on SECOND.

3D-CVF This is the official implementation of 3D-CVF: Generating Joint Camera and LiDAR Features Using Cross-View Spatial Feature Fusion for 3D Object

YecheolKim 97 Dec 20, 2022
SAAVN - Sound Adversarial Audio-Visual Navigation,ICLR2022 (In PyTorch)

SAAVN SAAVN Code release for paper "Sound Adversarial Audio-Visual Navigation,IC

YinfengYu 10 Aug 30, 2022
[NeurIPS 2021] Garment4D: Garment Reconstruction from Point Cloud Sequences

Garment4D [PDF] | [OpenReview] | [Project Page] Overview This is the codebase for our NeurIPS 2021 paper Garment4D: Garment Reconstruction from Point

Fangzhou Hong 112 Dec 23, 2022
Tutorial materials for Part of NSU Intro to Deep Learning with PyTorch.

Intro to Deep Learning Materials are part of North South University (NSU) Intro to Deep Learning with PyTorch workshop series. (Slides) Related materi

Hasib Zunair 9 Jun 08, 2022
This dlib-based facial login system

Facial-Login-System This dlib-based facial login system is a technology capable of matching a human face from a digital webcam frame capture against a

Mushahid Ali 3 Apr 23, 2022
High-performance moving least squares material point method (MLS-MPM) solver.

High-Performance MLS-MPM Solver with Cutting and Coupling (CPIC) (MIT License) A Moving Least Squares Material Point Method with Displacement Disconti

Yuanming Hu 2.2k Dec 31, 2022
One-line your code easily but still with the fun of doing so!

One-liner-iser One-line your code easily but still with the fun of doing so! Have YOU ever wanted to write one-line Python code, but don't have the sa

5 May 04, 2022
Vector.ai assignment

fabio-tests-nisargatman Low Level Approach: ###Tables: continents: id*, name, population, area, createdAt, updatedAt countries: id*, name, population,

Ravi Pullagurla 1 Nov 09, 2021
Official repository for HOTR: End-to-End Human-Object Interaction Detection with Transformers (CVPR'21, Oral Presentation)

Official PyTorch Implementation for HOTR: End-to-End Human-Object Interaction Detection with Transformers (CVPR'2021, Oral Presentation) HOTR: End-to-

Kakao Brain 114 Nov 28, 2022