Alpha-IoU: A Family of Power Intersection over Union Losses for Bounding Box Regression

Overview

Alpha-IoU: A Family of Power Intersection over Union Losses for Bounding Box Regression

YOLOv5 with alpha-IoU losses implemented in PyTorch.

Example results on the test set of PASCAL VOC 2007 using YOLOv5s trained by the vanilla IoU loss (top row) and the alpha-IoU loss with alpha=3 (bottom row). The alpha-IoU loss performs better than the vanilla IoU loss because it can localize objects more accurately (image 1 and 2), thus can detect more true positive objects (image 3 to 5) and fewer false positive objects (image 6 and 7).

Example results on the val set of MS COCO 2017 using YOLOv5s trained by the vanilla IoU loss (top row) and the alpha-IoU loss with alpha=3 (bottom row). The alpha-IoU loss performs better than the vanilla IoU loss because it can localize objects more accurately (image 1), thus can detect more true positive objects (image 2 to 5) and fewer false positive objects (image 4 to 7). Note that image 4 and 5 detect both more true positive and fewer false positive objects.

Citation

If you use our method, please consider citing:

@inproceedings{Jiabo_Alpha-IoU,
  author    = {He, Jiabo and Erfani, Sarah and Ma, Xingjun and Bailey, James and Chi, Ying and Hua, Xian-Sheng},
  title     = {Alpha-IoU: A Family of Power Intersection over Union Losses for Bounding Box Regression},
  booktitle = {NeurIPS},
  year      = {2021},
}

Modifications

This repository is a fork of ultralytics/yolov5, with an implementation of alpha-IoU losses while keeping the code as close to the original as possible.

Alpha-IoU Losses

Alpha-IoU losses can be configured in Line 131 of utils/loss.py, functionesd as 'bbox_alpha_iou'. The alpha values and types of losses (e.g., IoU, GIoU, DIoU, CIoU) can be selected in this function, which are defined in utils/general.py. Note that we should use a small constant epsilon to avoid torch.pow(0, alpha) or denominator=0.

Install

Python>=3.6.0 is required with all requirements.txt installed including PyTorch>=1.7:

$ git clone https://github.com/Jacobi93/Alpha-IoU
$ cd Alpha-IoU
$ pip install -r requirements.txt

Configurations

Configuration files can be found in data. We do not change either 'voc.yaml' or 'coco.yaml' used in the original repository. However, we could do more experiments. E.g.,

voc25.yaml # randomly use 25% PASCAL VOC as the training set
voc50.yaml # randomly use 50% PASCAL VOC as the training set

Code for generating different small training sets is in generate_small_sets.py. Code for generating different noisy labels is in generate_noisy_labels.py, and we should change the 'img2label_paths' function in utils/datasets.py accordingly.

Implementation Commands

For detailed installation instruction and network training options, please take a look at the README file or issue of ultralytics/yolov5. Following are sample commands we used for training and testing YOLOv5 with alpha-IoU, with more samples in instruction.txt.

python train.py --data voc.yaml --hyp hyp.scratch.yaml --cfg yolov5s.yaml --batch-size 64 --epochs 300 --device '0'
python test.py --data voc.yaml --img 640 --conf 0.001 --weights 'runs/train/voc_yolov5s_iou/weights/best.pt' --device '0'
python detect.py --source ../VOC/images/detect500 --weights 'runs/train/voc_yolov5s_iou/weights/best.pt' --conf 0.25

We can also randomly generate some images for detection and visualization results in generate_detect_images.py.

Pretrained Weights

Here are some pretrained models using the configurations in this repository, with alpha=3 in all experiments. Details of these pretrained models can be found in runs/train. All results are tested using 'weights/best.pt' for each experiment. It is a very simple yet effective method so that people is able to quickly apply our method to existing models following the 'bbox_alpha_iou' function in utils/general.py. Note that YOLOv5 has been updated for many versions and all pretrained models in this repository are obtained based on the YOLOv5 version 4.0, where details of all versions for YOLOv5 can be found. Researchers are also welcome to apply our method to other object detection models, e.g., Faster R-CNN, DETR, etc.

Owner
Jacobi(Jiabo He)
Jacobi(Jiabo He)
This is code to fit per-pixel environment map with spherical Gaussian lobes, using LBFGS optimization

Spherical Gaussian Optimization This is code to fit per-pixel environment map with spherical Gaussian lobes, using LBFGS optimization. This code has b

41 Dec 14, 2022
DALL-Eval: Probing the Reasoning Skills and Social Biases of Text-to-Image Generative Transformers

DALL-Eval: Probing the Reasoning Skills and Social Biases of Text-to-Image Generative Transformers Authors: Jaemin Cho, Abhay Zala, and Mohit Bansal (

Jaemin Cho 98 Dec 15, 2022
Official pytorch implement for “Transformer-Based Source-Free Domain Adaptation”

Official implementation for TransDA Official pytorch implement for “Transformer-Based Source-Free Domain Adaptation”. Overview: Result: Prerequisites:

stanley 54 Dec 22, 2022
Convolutional Neural Network for Text Classification in Tensorflow

This code belongs to the "Implementing a CNN for Text Classification in Tensorflow" blog post. It is slightly simplified implementation of Kim's Convo

Denny Britz 5.5k Jan 02, 2023
A study project using the AA-RMVSNet to reconstruct buildings from multiple images

3d-building-reconstruction This is part of a study project using the AA-RMVSNet to reconstruct buildings from multiple images. Introduction It is exci

17 Oct 17, 2022
A human-readable PyTorch implementation of "Self-attention Does Not Need O(n^2) Memory"

memory_efficient_attention.pytorch A human-readable PyTorch implementation of "Self-attention Does Not Need O(n^2) Memory" (Rabe&Staats'21). def effic

Ryuichiro Hataya 7 Dec 26, 2022
The original weights of some Caffe models, ported to PyTorch.

pytorch-caffe-models This repo contains the original weights of some Caffe models, ported to PyTorch. Currently there are: GoogLeNet (Going Deeper wit

Katherine Crowson 9 Nov 04, 2022
Official Pytorch implementation of Online Continual Learning on Class Incremental Blurry Task Configuration with Anytime Inference (ICLR 2022)

The Official Implementation of CLIB (Continual Learning for i-Blurry) Online Continual Learning on Class Incremental Blurry Task Configuration with An

NAVER AI 34 Oct 26, 2022
Dataset and Source code of paper 'Enhancing Keyphrase Extraction from Academic Articles with their Reference Information'.

Enhancing Keyphrase Extraction from Academic Articles with their Reference Information Overview Dataset and code for paper "Enhancing Keyphrase Extrac

15 Nov 24, 2022
Arxiv harvester - Poor man's simple harvester for arXiv resources

Poor man's simple harvester for arXiv resources This modest Python script takes

Patrice Lopez 5 Oct 18, 2022
MogFace: Towards a Deeper Appreciation on Face Detection

MogFace: Towards a Deeper Appreciation on Face Detection Introduction In this repo, we propose a promising face detector, termed as MogFace. Our MogFa

48 Dec 20, 2022
Official PyTorch implementation of Learning Intra-Batch Connections for Deep Metric Learning (ICML 2021) published at International Conference on Machine Learning

About This repository the official PyTorch implementation of Learning Intra-Batch Connections for Deep Metric Learning. The config files contain the s

Dynamic Vision and Learning Group 41 Dec 10, 2022
This is a file about Unet implemented in Pytorch

Unet this is an implemetion of Unet in Pytorch and it's architecture is as follows which is the same with paper of Unet component of Unet Convolution

Dragon 1 Dec 03, 2021
ResNEsts and DenseNEsts: Block-based DNN Models with Improved Representation Guarantees

ResNEsts and DenseNEsts: Block-based DNN Models with Improved Representation Guarantees This repository is the official implementation of the empirica

Kuan-Lin (Jason) Chen 2 Oct 02, 2022
The official homepage of the (outdated) COCO-Stuff 10K dataset.

COCO-Stuff 10K dataset v1.1 (outdated) Holger Caesar, Jasper Uijlings, Vittorio Ferrari Overview Welcome to official homepage of the COCO-Stuff [1] da

Holger Caesar 263 Dec 11, 2022
Latent Network Models to Account for Noisy, Multiply-Reported Social Network Data

VIMuRe Latent Network Models to Account for Noisy, Multiply-Reported Social Network Data. If you use this code please cite this article (preprint). De

6 Dec 15, 2022
A simple pytorch pipeline for semantic segmentation.

SegmentationPipeline -- Pytorch A simple pytorch pipeline for semantic segmentation. Requirements : torch=1.9.0 tqdm albumentations=1.0.3 opencv-pyt

petite7 4 Feb 22, 2022
Code for Multiple Instance Active Learning for Object Detection, CVPR 2021

Language: 简体中文 | English Introduction This is the code for Multiple Instance Active Learning for Object Detection, CVPR 2021. Installation A Linux pla

Tianning Yuan 269 Dec 21, 2022
Code and Experiments for ACL-IJCNLP 2021 Paper Mind Your Outliers! Investigating the Negative Impact of Outliers on Active Learning for Visual Question Answering.

Code and Experiments for ACL-IJCNLP 2021 Paper Mind Your Outliers! Investigating the Negative Impact of Outliers on Active Learning for Visual Question Answering.

Sidd Karamcheti 50 Nov 16, 2022
Another pytorch implementation of FCN (Fully Convolutional Networks)

FCN-pytorch-easiest Trying to be the easiest FCN pytorch implementation and just in a get and use fashion Here I use a handbag semantic segmentation f

Y. Dong 158 Dec 21, 2022