#144144th of 168 in ML frameworks and libraries
weighted-boxes-fusion
Python library for ensembling bounding boxes from multiple object detection models
- Stars
- 1.8k
- License
- MIT
- Last commit
- Oct 2026
- Last release
- Mar 2022
- Language
- Python
1,811 on 21 Apr 20261,833 on 10 Nov 2026
From GitHub's daily star counts.
Checked on GitHub.
Overview
ensemble-boxes is a Python package that merges predictions from several object detection models using NMS, Soft-NMS, Non-maximum weighted and Weighted Boxes Fusion (WBF). WBF also has 3D, 1D line-segment, rotated-box and quadrangle variants. Inputs are normalized box coordinates, scores and labels per model, with optional per-model weights.
Who it is for: ML engineers ensembling object detection model outputs
Strengths
- Four ensembling methods behind one call signature: nms, soft_nms, non_maximum_weighted, weighted_boxes_fusion
- WBF variants for 3D, 1D segments, rotated and quadrangle boxes
- Installs with pip; requires only Numpy and Numba
- Benchmarks provided for Open Images, COCO and an NLP dataset
Weaknesses
- Boxes must be pre-normalized to [0, 1] in x1, y1, x2, y2 order
- Rotated boxes need the same scale on both axes, or a square image
- Last tagged release is from 2022-03-24
- Library only; no model inference or training included
What it needs
- no GPU
- Needs Numpy, Numba
- Models: object detection box outputs (boxes, scores, labels)
Quick answers
Is weighted-boxes-fusion open source?
Yes. Its code is on GitHub under the MIT license.
Does weighted-boxes-fusion need a GPU?
No, it runs without one.
Also in ML frameworks and libraries
See all 168| Rank | Project | Score |
|---|---|---|
| 1 | transformersPython library defining open models for text, vision, audio and multimodal tasks | 94 out of 100 |
| 2 | pytorchPython tensor library with GPU support and tape-based autograd | 93 out of 100 |
| 3 | ultralyticsPython library and CLI for training and running YOLO vision models | 93 out of 100 |
| #4 | pytorch-image-modelsPyTorch library of image models, pretrained weights and training scripts | 93 out of 100 |
| #5 | supervisionPython toolkit for loading, annotating and counting computer vision detections | 92 out of 100 |