graviti
Products
Resources
About us
FSS-1000
2D Polygon
Common
|...
License: Unknown

Overview

FSS-1000 is a few-shot segmentation dataset where the number of annotated training examples
are limited to 5 only. It consists of 1000 object classes with pixelwise annotation of ground-truth
segmentation. Unique in FSS-1000, our dataset contains significant number of objects that have
never been seen or annotated in previous datasets, such as tiny daily objects, merchandise,
cartoon characters, logos, etc.

Data Collection

Object Classes

We first referred to the classes in ILSVRC in our choice of object categories for FSS1000.
Consequently, FSS-1000 has 584 classes out of its 1,000 classes overlap with the classes in
the ILSVRC dataset. We find ILSVRC dataset heavily biases toward animals, both in terms of
the distribution of categories and number of images. Therefore, we fill in the other 486 by
new classes unseen in any existing datasets. Specifically, we include more daily objects so
that network models trained on FSS-1000 can learn from diverse artificial and manmade objects/features
in addition to natural and organic objects/features where the latter was emphasized by existing
large-scale datasets.

Raw Images

To avoid bias, the raw images were retrieved by querying object keywords
on three different Internet search engines, namely, Google, Bing and Yahoo. We downloaded the
first 100 results returned (or less if less than 100 images were returned) from a given search
engine. No special criteria or assumption was used to select the candidates, however, due to
the bias of Internet search engines, a large number of the images returned contain a single
object photographed with sharp focus. In the final step, we intentionally included some images
with a relatively small object, multiple objects or other objects in the background to balance
the easy and hard examples of the dataset. Images with aspect ratio larger than 2 or smaller
than 0.5 were excluded. Since all images and their segmentation maps were to be resized to
224×224, bad aspect ratio would destroy important geometric properties after the resize operation.
For the same reason, images with height or width less than 224 pixels were discarded because
they would trigger upsampling which would affect the image quality after resizing.

Pixelwise Segmentation Annotation

We used
Photoshop’s “quick selection" tool which allows users to loosely select an object automatically,
and refined or corrected the selected area to produce the desired segmentation.

Instruction

Prerequisites

  • Linux or macOS
  • Python 3
  • CPU or NVIDIA GPU + CUDA CuDNN
  • PyTorch 0.4+

Testing

First, download pretrained model here.

python autolabel.py -sd imgs/example/support -td imgs/example/query
  • Set option -sd to the support directory and the script will input them as support set.
  • Set option -td to the path of your query images.
  • Results will be saved under ./results

Testing your own data

  • Label 5 support images following the format in imgs/example/support/.
  • Set your support and query path accordingly.

Training

Arrange the dataset as described in get_oneshot_batch() in training.py, then run

python training.py

Citation

@article{FSS1000,
Author = {Xiang Li and Tianhan Wei and Yau Pun Chen and Yu-Wing Tai and Chi-Keung Tang},
Title = {FSS-1000: A 1000-Class Dataset for Few-Shot Segmentation},
Year = {2020},
Journal = {CVPR},
}
Data Summary
Type
Image,
Amount
--
Size
649.63MB
Provided by
HKUSTCV
The Hong Kong University of Science and Technology (HKUST) is a public research university in Clear Water Bay, Hong Kong. Founded in 1991 by the British Hong Kong Government, it was the territory's third institution to be granted university status.
| Amount -- | Size 649.63MB
FSS-1000
2D Polygon
Common
License: Unknown

Overview

FSS-1000 is a few-shot segmentation dataset where the number of annotated training examples
are limited to 5 only. It consists of 1000 object classes with pixelwise annotation of ground-truth
segmentation. Unique in FSS-1000, our dataset contains significant number of objects that have
never been seen or annotated in previous datasets, such as tiny daily objects, merchandise,
cartoon characters, logos, etc.

Data Collection

Object Classes

We first referred to the classes in ILSVRC in our choice of object categories for FSS1000.
Consequently, FSS-1000 has 584 classes out of its 1,000 classes overlap with the classes in
the ILSVRC dataset. We find ILSVRC dataset heavily biases toward animals, both in terms of
the distribution of categories and number of images. Therefore, we fill in the other 486 by
new classes unseen in any existing datasets. Specifically, we include more daily objects so
that network models trained on FSS-1000 can learn from diverse artificial and manmade objects/features
in addition to natural and organic objects/features where the latter was emphasized by existing
large-scale datasets.

Raw Images

To avoid bias, the raw images were retrieved by querying object keywords
on three different Internet search engines, namely, Google, Bing and Yahoo. We downloaded the
first 100 results returned (or less if less than 100 images were returned) from a given search
engine. No special criteria or assumption was used to select the candidates, however, due to
the bias of Internet search engines, a large number of the images returned contain a single
object photographed with sharp focus. In the final step, we intentionally included some images
with a relatively small object, multiple objects or other objects in the background to balance
the easy and hard examples of the dataset. Images with aspect ratio larger than 2 or smaller
than 0.5 were excluded. Since all images and their segmentation maps were to be resized to
224×224, bad aspect ratio would destroy important geometric properties after the resize operation.
For the same reason, images with height or width less than 224 pixels were discarded because
they would trigger upsampling which would affect the image quality after resizing.

Pixelwise Segmentation Annotation

We used
Photoshop’s “quick selection" tool which allows users to loosely select an object automatically,
and refined or corrected the selected area to produce the desired segmentation.

Instruction

Prerequisites

  • Linux or macOS
  • Python 3
  • CPU or NVIDIA GPU + CUDA CuDNN
  • PyTorch 0.4+

Testing

First, download pretrained model here.

python autolabel.py -sd imgs/example/support -td imgs/example/query
  • Set option -sd to the support directory and the script will input them as support set.
  • Set option -td to the path of your query images.
  • Results will be saved under ./results

Testing your own data

  • Label 5 support images following the format in imgs/example/support/.
  • Set your support and query path accordingly.

Training

Arrange the dataset as described in get_oneshot_batch() in training.py, then run

python training.py

Citation

@article{FSS1000,
Author = {Xiang Li and Tianhan Wei and Yau Pun Chen and Yu-Wing Tai and Chi-Keung Tang},
Title = {FSS-1000: A 1000-Class Dataset for Few-Shot Segmentation},
Year = {2020},
Journal = {CVPR},
}
0
Start building your AI now
graviti
wechat-QR
Long pressing the QR code to follow wechat official account

Copyright@Graviti