Chair Segments: A Compact Benchmark for the Study of Object Segmentation
Over the years, datasets and benchmarks have had an outsized influence on the design of novel algorithms. In this paper, we introduce ChairSegments, a novel and compact semi-synthetic dataset for object segmentation. We also show empirical findings in transfer learning that mirror recent findings for image classification. We particularly show that models that are fine-tuned from a pretrained set of weights lie in the same basin of the optimization landscape. ChairSegments consists of a diverse set of prototypical images of chairs with transparent backgrounds composited into a diverse array of backgrounds. We aim for ChairSegments to be the equivalent of the CIFAR-10 dataset but for quickly designing and iterating over novel model architectures for segmentation. On Chair Segments, a U-Net model can be trained to full convergence in only thirty minutes using a single GPU. Finally, while this dataset is semi-synthetic, it can be a useful proxy for real data, leading to state-of-the-art accuracy on the Object Discovery dataset when used as a source of pretraining.
- Python 3
- Pytorch > 1.0
- torchVision
sh download_data.sh
# Start training with:
# ChairSegment - parameters
python main.py --lr=1e-3 --arch=unet --optimizer=Adam --epochs=20
python main.py --lr=1e-4 --arch=fcnvgg16 --optimizer=SGD --epochs=50
python main.py --lr=1e-4 --arch=fcnresnet50 --optimizer=SGD --epochs=100 --momentum=0.9 --weight_decay=1e-5
python main.py --lr=1e-6 --arch=fcnresnet101 --optimizer=RMSprop --epochs=100 --momentum=0.9 --weight_decay=1e-7
| Model | Prec. | IoU. | Dice |
|---|---|---|---|
| Unet | 97.18% | 85.08% | 91.25% |
| FCN-VGG-16 | 91.73% | 61.09% | 74.09% |
| FCN-ResNet-50 | 92.04% | 60.19% | 72.58% |
| FCN-ResNet-101 | 92.19% | 61.62% | 73.96% |
Report: https://arxiv.org/abs/2012.01250
@misc{pintoalva2020chair,
title={Chair Segments: A Compact Benchmark for the Study of Object Segmentation},
author={Leticia Pinto-Alva and Ian K. Torres and Rosangel Garcia and Ziyan Yang and Vicente Ordonez},
year={2020},
eprint={2012.01250},
archivePrefix={arXiv},
primaryClass={cs.CV}
}