p3-ims-obd-ensemble

BoostCamp AI Tech P Stage stage3 Image Segmentation

List

• Segmentation
  * JooYoung_dev
  * HyeongMing_dev
  * JunCheol_dev
  * Kimin_dev
  * MinJung_dev
  * NuRee_dev
• Detection
  * JooYoung_dev
  * HyeongMing_dev
  * JunCheol_dev
  * Kimin_dev
  * MinJung_dev
  * NuRee_dev

Segmentation

JooYoung_dev

You can see this branch at main too.

$> tree -d
.
├── segmentaions
|     ├── config
|     ├── data
|     ├── losses
|     ├── models
|     ├── scheduler
|     ├── utils
│     └── ...
└── detection
      └── ...

• config
  * efficient b2/b6 + unet++
  * efficient b4/b6 + deepLabV3+
  * effunext
  * resnext101 + upernet
  * resnext50 + deepLabV3+
  * unext
• losses
  * dice_ce_loss
  * soft_ce_loss
• scheduler * customcosine

HyeongMin_dev

$> tree -d
.
├── Segmentaion
│     └── ...
└── detection
      └── ...

• model: resnext50_32x4d + DeepLabV3+

However, because used smp, you can use others in smp

JunCheol_dev

$> tree -d
.
├── segmentaion
|     ├── ppt_paper
│     └── ...
└── detection
      └── ...

• ppt_paper
  * segmentation_survey_2020.pdf
  * Rethinking Pre-training and Self-training.pptx
  * Imgage Segmentation.pptx
  * EDA.pptx
• model: emanet

Kimin_dev

$> tree -d
.
├── segmentaions
│     └── ...
└── detection
      └── ...

• TTA_CRF
• psheudo_train

MinJung_dev

$> tree -d
.
├── seg
|     ├── Tools
|     ├── dataset
|     ├── lib
|     ├── preprocess
│     └── ...
└── detection
      └── ...

• lib
  * asgnet
  * hrdnet

NuRee_dev

$> tree -d
.
├── segmentation
│     └── ...
└── detection
      └── ...

• augmentaion test
• Resnext50 + DeepLabV3+

Object Detection

JooYoung_dev

$> tree -d
.
├── segmentaions
│     └── ...
└── detection
      ├── UniverseNet-master
      └── ...

mmdet

• universNet

HyeongMin_dev

$> tree -d
.
├── segmentaion
│     └── ...
└── ObjectDetection
      ├── Swin-Transformer-Object-Detection
      └── ...

mmdet

• detectoRS + ResNeXt101

JunCheol_dev

$> tree -d
.
├── segmentaion
│     └── ...
└── ObjectDetection
      ├── Swin-Transformer-Object-Detection
      └── ...

mmdet

• Swin-S

Kimin_dev

$> tree -d
.
├── segmentaions
│     └── ...
└── detection
      ├── efficientdet-naive-pytorch
      ├── faster_rcnn-naive-pytorch
      └── vfnetx-mmdet

naive model

• efficientdet
• faster_rcnn

mmdet

• vfnetx

MinJung_dev

$> tree -d
.
├── seg
│     └── ...
└── detection
      ├── Swin-Transformer-Object-Detection-master
      ├── mmdet
      └── augmentaion

mmdet

• Swin-T
• efficienet b3 - nas-fpn - cascade rcnn

NuRee_dev

$> tree -d
.
├── segmentation
│     └── ...
└── detection
      ├── Swin-Transformer-Object-Detection-master
      └── ...

mmdet

• Swin-T