BoxInst: High-Performance Instance Segmentation with Box Annotations

By Zhi Tian, Chunhua Shen, Xinlong Wang, Hao Chen. 2020

324 citations recorded by the source corpus.

We present a high-performance method that can achieve mask-level instance segmentation with only bounding-box annotations for training. While this setting has been studied in the literature, here we show significantly stronger performance with a simple design (e.g., dramatically improving previous best reported mask AP of 21.1% [13] to 31.6% on the COCO dataset). Our core idea is to redesign the loss of learning masks in instance segmentation, with no modification to the segmentation network itself. The new loss functions can supervise the mask training without relying on mask annotations. This is made possible with two loss terms, namely, 1) a surrogate term that minimizes the discrepancy between the projections of the ground-truth box and the predicted mask; 2) a pairwise loss that can exploit the prior that proximal pixels with similar colors are very likely to have the same category label.Experiments demonstrate that the redesigned mask loss can yield surprisingly high-quality instance masks with only box annotations. For example, without using any mask annotations, with a ResNet-101 backbone and 3× training schedule, we achieve 33.2% mask AP on COCO test-dev split (vs. 39.1% of the fully supervised counterpart). Our excellent experiment results on COCO and Pascal VOC indicate that our method dramatically narrows the performance gap between weakly and fully supervised instance segmentation.Code is available at: https://git.io/AdelaiDet

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Computer Vision and Pattern Recognition

Citations in the corpus

A Simple Framework for Open-Vocabulary Segmentation and Detection (2023, methodology): “Prominent methods design teacher models or weak supervision loss, like BoxInst [42], Box2Mask [30], DiscoBox [25] and Mask AutoLabelers [24].”
Fine-Grained Egocentric Hand-Object Segmentation: Dataset, Model, and Applications (2022, methodology): “We compare with 100-DOH by training a weakly supervised segmentation model, BoxInst [70], which learns to segment objects given bounding box annotation.”
DanceTrack: Multi-Object Tracking in Uniform Appearance and Diverse Motion (2021, background): “4 that DanceTrack only provides bounding box annotation, for now, there would be several interesting future works: (1) extending its annotation modalities, (2) using weaklysupervised learning [34,38,49] to estimate other modalities, (3) using transfer learning and domain adaptation [1, 5, 22] to tra”
Open-Vocabulary Instance Segmentation via Robust Cross-Modal Pseudo-Labeling (2021, background): “Given bounding box annotations, [15, 20, 21, 30, 31] exploit pixelwise similarity to infer object masks while [18, 19, 32, 33] learn to share mask knowledge between mask and box supervision to enhance performances.”
Transformer-Based Visual Segmentation: A Survey (2023, background): “Video K-Net [172] achieves the best results on the KITTI-STEP dataset.”
Box-supervised Instance Segmentation with Level Set Evolution (2022, methodology): “4% AP, which is higher than BoxInst [44] by 0.”
Pointly-Supervised Instance Segmentation (2021, methodology): “The recent BoxInst model [53] outperforms previous weakly-supervised approaches with box supervision but achieves only 85% of its fully-supervised counterpart performance on COCO.”
DiscoBox: Weakly Supervised Instance Segmentation and Semantic Correspondence from Box Supervision (2021, methodology): “9 BoxInst [49] ResNet-101-DCN One-stage CondInst Box - 35.”
Fully Convolutional Networks for Panoptic Segmentation With Point-Based Supervision (2021, methodology): “Therefore, box-level annotation [10], [11], [12] and point-based manners [13], [14], [15] are utilized ar X iv :2 10 8.”
RSPrompter: Learning to Prompt for Remote Sensing Instance Segmentation Based on Visual Foundation Model (2023, methodology): “We also take into account single-stage methods, such as SOLOv2 [89], CondInst [28], BoxInst [91], and Mask2Former [30].”