Dataset_Setup_Instructions.md

Computer Vision Datasets Setup

SuperGradients provides multiple Datasets implementations.

Classification Datasets

Cifar10

Supports download

from super_gradients.training.datasets import Cifar10
dataset = Cifar10(..., download=True)

Imagenet

1. Download imagenet dataset:

• https://image-net.org/download.php

2. Unzip:

    Imagenet
     ├──train
     │  ├──n02093991
     │  │   ├──n02093991_1001.JPEG
     │  │   ├──n02093991_1004.JPEG
     │  │   └──...
     │  ├──n02093992
     │  └──...
     └──val
        ├──n02093991
        ├──n02093992
        └──...

3. Instantiate the dataset:

from super_gradients.training.datasets import ImageNetDataset
train_set = ImageNetDataset(root='.../Imagenet/train', ...)
valid_set = ImageNetDataset(root='.../Imagenet/val', ...)

Detection Datasets

Coco

1. Download coco dataset:

• annotations: http://images.cocodataset.org/annotations/annotations_trainval2017.zip
• train2017: http://images.cocodataset.org/zips/train2017.zip
• val2017: http://images.cocodataset.org/zips/val2017.zip

2. Unzip and organize it as below:

    coco
    ├── annotations
    │      ├─ instances_train2017.json
    │      ├─ instances_val2017.json
    │      └─ ...
    └── images
        ├── train2017
        │   ├─ 000000000001.jpg
        │   └─ ...
        └── val2017
            └─ ...

3. Instantiate the dataset:

from super_gradients.training.datasets import COCODetectionDataset
train_set = COCODetectionDataset(data_dir='.../coco', subdir='images/train2017', json_file='instances_train2017.json', ...)
valid_set = COCODetectionDataset(data_dir='.../coco', subdir='images/val2017', json_file='instances_val2017.json', ...)

PascalVOC 2007 & 2012

Supports download

from super_gradients.training.datasets import PascalVOCDetectionDataset
train_set = PascalVOCDetectionDataset(download=True, ...)

Dataset Structure:

    Dataset structure:
        ├─images
        │   ├─ train2012
        │   ├─ val2012
        │   ├─ VOCdevkit
        │   │    ├─ VOC2007
        │   │    │  ├──JPEGImages
        │   │    │  ├──SegmentationClass
        │   │    │  ├──ImageSets
        │   │    │  ├──ImageSets/Segmentation
        │   │    │  ├──ImageSets/Main
        │   │    │  ├──ImageSets/Layout
        │   │    │  ├──Annotations
        │   │    │  └──SegmentationObject
        │   │    └──VOC2012
        │   │       ├──JPEGImages
        │   │       ├──SegmentationClass
        │   │       ├──ImageSets
        │   │       ├──ImageSets/Segmentation
        │   │       ├──ImageSets/Main
        │   │       ├──ImageSets/Action
        │   │       ├──ImageSets/Layout
        │   │       ├──Annotations
        │   │       └──SegmentationObject
        │   ├─train2007
        │   ├─test2007
        │   └─val2007
        └─labels
            ├─train2012
            ├─val2012
            ├─train2007
            ├─test2007
            └─val2007

Yolo/Darknet format

1. Download your dataset (can be from https://roboflow.com/universe)

2. You should have a structure similar to this.

     data_dir
     └── train/test/val
         ├── images
         │      ├─ 0001.jpg
         │      ├─ 0002.jpg
         │      └─ ...
         └── labels
                ├─ 0001.txt
                ├─ 0002.txt
                └─ ...

Note: train/test/val folders are not required, any folder structure is supported.

3. Instantiate the dataset:

from super_gradients.training.datasets import YoloDarknetFormatDetectionDataset
data_set = YoloDarknetFormatDetectionDataset(data_dir='<path-to>/data_dir', images_dir="<train/test/val>/images", labels_dir="<train/test/val>/labels", classes=["<to-fill>"])

Segmentation Datasets

Cityscapes

1. Download dataset:

• a. Cityscapes dataset:
  • gtFine: https://www.cityscapes-dataset.com/file-handling/?packageID=1
  • leftImg8bit: https://www.cityscapes-dataset.com/file-handling/?packageID=3
• b. metadata folder: https://deci-pretrained-models.s3.amazonaws.com/cityscape_lists.zip

2. a. Unzip and organize cityscapes dataset as below:

  root_dir (in recipe default to /data/cityscapes)
      ├─── gtFine
      │       ├── test
      │       │     ├── berlin
      │       │     │   ├── berlin_000000_000019_gtFine_color.png
      │       │     │   ├── berlin_000000_000019_gtFine_instanceIds.png
      │       │     │   └── ...
      │       │     ├── bielefeld
      │       │     │   └── ...
      │       │     └── ...
      │       ├─── train
      │       │     └── ...
      │       └─── val
      │             └── ...
      └─── leftImg8bit
              ├── test
              │     └── ...
              ├─── train
              │     └── ...
              └─── val
                    └── ...

2. b. Unzip and organize metadata folder as below:

  lists
      ├── labels.csv
      ├── test.lst
      ├── train.lst
      ├── trainval.lst
      ├── val.lst
      └── auto_labelling.lst

2. c. Move Metadata folder to the Cityscapes folder

  root_dir (in recipe default to /data/cityscapes)
      ├─── gtFine
      │      └── ...
      ├─── leftImg8bit
      │      └── ...
      └─── lists
             └── ...

3. Instantiate the dataset:

from super_gradients.training.datasets import CityscapesDataset

train_set = CityscapesDataset(root_dir='.../root_dir', list_file='lists/train.lst', labels_csv_path='lists/labels.csv', ...)

4. AutoLabelling dataset [Optional]

Cityscapes AutoLabelled dataset were introduced by NVIDIA research group in the paper: "Hierarchical Multi-Scale Attention for Semantic Segmentation".

AutoLabelled refer to the refinement of the Cityscapes coarse data and pseudo labels generation using their suggested Hierarchical multi-scale attention model.

• To download the AutoLabelled labels please refer to the original repo. Unzip and rename the folder to AutoLabelling as described bellow.

• Download the coarse RGB images from cityscapes official site, leftImg8bit_train_extra: https://www.cityscapes-dataset.com/file-handling/?packageID=4

  root_dir (in recipe default to /data/cityscapes)
      ├─── gtFine
      │       ├── test
      │       │     └── ...
      │       ├─── train
      │       │     └── ...
      │       └─── val
      │             └── ...
      ├─── leftImg8bit
      │       ├── test
      │       │     └── ...
      │       ├─── train
      │       │     └── ...
      │       └─── val
      │             └── ...
      ├─── AutoLabelling
      │       └─── train_extra
      │             └── ...
      └─── leftImg8bit
              └─── train_extra
                    └── ...

Coco

1. Download coco dataset:

• annotations: http://images.cocodataset.org/annotations/annotations_trainval2017.zip
• train2017: http://images.cocodataset.org/zips/train2017.zip
• val2017: http://images.cocodataset.org/zips/val2017.zip

2. Unzip and organize it as below:

    coco
    ├── annotations
    │      ├─ instances_train2017.json
    │      ├─ instances_val2017.json
    │      └─ ...
    └── images
        ├── train2017
        │   ├─ 000000000001.jpg
        │   └─ ...
        └── val2017
            └─ ...

3. Instantiate the dataset:

from super_gradients.training.datasets import CoCoSegmentationDataSet
train_set = CoCoSegmentationDataSet(data_dir='.../coco', subdir='images/train2017', json_file='instances_train2017.json', ...)
valid_set = CoCoSegmentationDataSet(data_dir='.../coco', subdir='images/val2017', json_file='instances_val2017.json', ...)

Pascal VOC 2012

1. Download pascal datasets:

• VOC 2012: http://host.robots.ox.ac.uk/pascal/VOC/voc2012/VOCtrainval_11-May-2012.tar

2. Unzip and organize it as below:

   pascal_voc_2012
       └──VOCdevkit
             └──VOC2012
                ├──JPEGImages
                ├──SegmentationClass
                ├──ImageSets
                │    ├──Segmentation
                │    │   └── train.txt
                │    ├──Main
                │    ├──Action
                │    └──Layout
                ├──Annotations
                └──SegmentationObject

3. Instantiate the dataset:

from super_gradients.training.datasets import PascalVOC2012SegmentationDataSet

train_set = PascalVOC2012SegmentationDataSet(
     root='.../pascal_voc_2012',
     list_file='VOCdevkit/VOC2012/ImageSets/Segmentation/train.txt',
     samples_sub_directory='VOCdevkit/VOC2012/JPEGImages',
     targets_sub_directory='VOCdevkit/VOC2012/SegmentationClass',
     ...
 )
valid_set = PascalVOC2012SegmentationDataSet(
     root='.../pascal_voc_2012',
     list_file='VOCdevkit/VOC2012/ImageSets/Segmentation/val.txt',
     samples_sub_directory='VOCdevkit/VOC2012/JPEGImages',
     targets_sub_directory='VOCdevkit/VOC2012/SegmentationClass',
     ...
 )

Pascal AUG 2012

1. Download pascal dataset

• AUG 2012: https://www2.eecs.berkeley.edu/Research/Projects/CS/vision/grouping/semantic_contours/benchmark.tgz

2. Unzip and organize it as below:

   pascal_voc_2012
       └──VOCaug
           ├── aug.txt
           └── dataset
                 ├──inst
                 ├──img
                 └──cls

3. Instantiate the dataset:

from super_gradients.training.datasets import PascalAUG2012SegmentationDataSet

train_set = PascalAUG2012SegmentationDataSet(
     root='.../pascal_voc_2012',
     list_file='VOCaug/dataset/aug.txt',
     samples_sub_directory='VOCaug/dataset/img',
     targets_sub_directory='VOCaug/dataset/cls',
     ...
 )

NOTE: this dataset is only available for training. To test, please use PascalVOC2012SegmentationDataSet.

Pascal AUG & VOC 2012

1. Download pascal datasets:

• VOC 2012: http://host.robots.ox.ac.uk/pascal/VOC/voc2012/VOCtrainval_11-May-2012.tar
• AUG 2012: https://www2.eecs.berkeley.edu/Research/Projects/CS/vision/grouping/semantic_contours/benchmark.tgz

2. Unzip and organize it as below:

   pascal_voc_2012
       ├─VOCdevkit
       │ └──VOC2012
       │    ├──JPEGImages
       │    ├──SegmentationClass
       │    ├──ImageSets
       │    │    ├──Segmentation
       │    │    │   └── train.txt
       │    │    ├──Main
       │    │    ├──Action
       │    │    └──Layout
       │    ├──Annotations
       │    └──SegmentationObject
       └──VOCaug
           ├── aug.txt
           └── dataset
                 ├──inst
                 ├──img
                 └──cls

3. Instantiate the dataset:

from super_gradients.training.datasets import PascalVOCAndAUGUnifiedDataset
train_set = PascalVOCAndAUGUnifiedDataset(root='.../pascal_voc_2012', ...)

NOTE: this dataset is only available for training. To test, please use PascalVOC2012SegmentationDataSet.

Supervisely Persons

1. Download supervisely dataset:

• https://deci-pretrained-models.s3.amazonaws.com/supervisely-persons.zip

2. Unzip:

   supervisely-persons
    ├──images
    │    ├──image-name.png
    │    └──...
    ├──images_600x800
    │    ├──image-name.png
    │    └──...
    ├──masks
    └──masks_600x800

3. Instantiate the dataset:

from super_gradients.training.datasets import SuperviselyPersonsDataset
train_set = SuperviselyPersonsDataset(root_dir='.../supervisely-persons', list_file='train.csv', ...)
valid_set = SuperviselyPersonsDataset(root_dir='.../supervisely-persons', list_file='val.csv', ...)

NOTE: this dataset is only available for training. To test, please use PascalVOC2012SegmentationDataSet.

Pose Estimation Datasets

COCO 2017

1. Download coco dataset:

• annotations: http://images.cocodataset.org/annotations/annotations_trainval2017.zip
• train2017: http://images.cocodataset.org/zips/train2017.zip
• val2017: http://images.cocodataset.org/zips/val2017.zip

2. Unzip and organize it as below:

    coco
    ├── annotations
    │      ├─ person_keypoints_train2017.json
    │      ├─ person_keypoints_val2017.json
    │      └─ ...
    └── images
        ├── train2017
        │   ├─ 000000000001.jpg
        │   └─ ...
        └── val2017
            └─ ...

3. Instantiate the dataset:

from super_gradients.training.datasets import COCOKeypointsDataset
train_set = COCOKeypointsDataset(data_dir='.../coco', images_dir='images/train2017', json_file='annotations/instances_train2017.json', ...)
valid_set = COCOKeypointsDataset(data_dir='.../coco', images_dir='images/val2017', json_file='annotations/instances_val2017.json', ...)