Skip to content

Latest commit

 

History

5 Commits

Folders and files

NameName
Last commit message
Last commit date
 
 
 
 
 
 
 
 
 
 

Repository files navigation

NYCU Visual Recognition Using Deep Learning 2025 Spring Final Project

Sartorius Cell Instance Segmentation

StudentID: 313540009 Name: Anna Kompan (安娜)

Introduction

Task: Cell Instance Segmentation Source: Kaggle Competition Content:

  • train.csv — IDs and masks for all training objects.
  • sample_submission.csv - a sample submission file in the correct format
  • train - train images in PNG format
  • test - test images in PNG format.
  • train_semi_supervised - unlabeled images offered for a semi-supervised approach.
  • LIVECell_dataset_2021 - A mirror of the data from the LIVECell dataset. LIVECell is the predecessor dataset to this competition.

Install Dependencies

pip install numpy pandas matplotlib opencv-python torch torchvision tqdm

Recommend using Conda for version cotrol Python version used is 3.11.11

Training & Evaluation

Preprocessing:

  • Converts RLE masks to binary masks
  • Groups masks by image ID

Train Images and Masks

Train_data

Augmentation:

  • Customized Vertical/Horizontal Flip (for masks and images)
  • Normalize
  • ToTensor

Model Training:

  • Download pretrained weights maskrcnn_resnet50_fpn
  • Replace classification and mask head for dataset number of classes + 1 (background)
  • Train 50 epochs
  • Learning rate scheduler every 5 epochs
  • Optimizer (SGD with Momentum and Weight decay)
  • (Optional) Add early stopping

Training and Validation Loss Curve

Loss_Curve

Evaluation:

  • Evaluate on test dataset
  • Save submission.csv with columns(id, predicted)

Evaluation Results

Evaluation

Performance Snapshot

Performance snapshot

About

Visual Recognition using Deep Learning Final Project (Kaggle Competition)

Resources

Stars

0 stars

Watchers

0 watching

Forks

Releases

Packages

Contributors

Languages