Seach Losses of our paper 'Loss Function Discovery for Object Detection via Convergence-Simulation Driven Search', accepted by ICLR 2021.

Overview

CSE-Autoloss

Designing proper loss functions for vision tasks has been a long-standing research direction to advance the capability of existing models. For object detection, the well-established classification and regression loss functions have been carefully designed by considering diverse learning challenges (e.g. class imbalance, hard negative samples, and scale variances). Inspired by the recent progress in network architecture search, it is interesting to explore the possibility of discovering new loss function formulations via directly searching the primitive operation combinations. So that the learned losses not only fit for diverse object detection challenges to alleviate huge human efforts, but also have better alignment with evaluation metric and good mathematical convergence property. Beyond the previous auto-loss works on face recognition and image classification, our work makes the first attempt to discover new loss functions for the challenging object detection from primitive operation levels and finds the searched losses are insightful. We propose an effective convergence-simulation driven evolutionary search algorithm, called CSE-Autoloss, for speeding up the search progress by regularizing the mathematical rationality of loss candidates via two progressive convergence simulation modules: convergence property verification and model optimization simulation. The best-discovered loss function combinations CSE-Autoloss-A and CSE-Autoloss-B outperform default combinations (Cross-entropy/Focal loss for classification and L1 loss for regression) by 1.1% and 0.8% in terms of mAP for two-stage and one-stage detectors on COCO respectively.

The repository contains the demo training scripts for the best-searched loss combinations of our paper (ICLR2021) Loss Function Discovery for Object Detection via Convergence-Simulation Driven Search.

Installation

Please refer to get_started.md for installation.

Get Started

Please see get_started.md for the basic usage of MMDetection.

Searched Loss

Two-Stage Best-Discovered Loss

CSE_Autoloss_A_cls='Neg(Dot(Mul(Y,Add(1,Sin(Z))),Log(Softmax(X))))'

CSE_Autoloss_A_reg='Add(1,Neg(Add(Div(I,U),Neg(Div(Add(E,Neg(Add(I,2))),E)))))'

One-Stage Best-Discovered Loss

CSE_Autoloss_B_cls='Neg(Add(Mul(Q,Mul(Add(1,Serf(Sig(NY))),Log(Sig(X)))),Mul(Add(Sgdf(X),Neg(Q)),Mul(Add(Add(1,Neg(Q)),Neg(Add(1,Neg(Sig(X))))),Log(Add(1,Neg(Sig(X))))))))'

CSE_Autoloss_B_reg='Neg(Div(Add(Div(Neg(Add(Neg(E),Add(1,I))),Neg(Add(3,Add(2,U)))),Add(Div(E,E),Div(Neg(E),Neg(1)))),Neg(Add(Div(Neg(Add(U,Div(I,1))),Neg(3)),Neg(E)))))'

[1] u, i, e, w indicate union, intersection, enclose and intersection-over-union (IoU) between bounding box prediction and groundtruth. x, y are for class prediction and label.
[2] dot is for dot product, erf is for scaled error function, gd is for scaled gudermannian function. Please see more details about "S"-shaped curve at wiki.

Performance

Performance for COCO val are as follows.

Detector Loss Bbox mAP Command
Faster R-CNN R50 CSE-Autoloss-A 38.5% Link
Faster R-CNN R101 CSE-Autoloss-A 40.2% Link
Cascade R-CNN R50 CSE-Autoloss-A 40.5% Link
Mask R-CNN R50 CSE-Autoloss-A 39.1% Link
FCOS R50 CSE-Autoloss-B 39.6% Link
ATSS R50 CSE-Autoloss-B 40.5% Link

[1] We replace the centerness_target in FCOS and ATSS to the IoU between bbox_pred and bbox_target. Please see more details at fcos_head.py and atss_head.py.

[2] For the search loss combinations, loss_bbox weight for ATSS sets to 1 (instead of 2). Please see more details here.

Quick start to train the model with searched/default loss combinations

# cls - classification, reg - regression

# Train with searched classification loss and searched regression loss
python -m torch.distributed.launch --nproc_per_node=$GPUS --master_port=$PORT ./tools/train.py $CONFIG --loss_cls $SEARCH_CLS_LOSS --loss_reg $SEARCH_REG_LOSS --launcher pytorch;

# Train with searched classification loss and default regression loss
python -m torch.distributed.launch --nproc_per_node=$GPUS --master_port=$PORT ./tools/train.py $CONFIG --loss_cls $SEARCH_CLS_LOSS --launcher pytorch;

# Train with default classification loss and searched regression loss
python -m torch.distributed.launch --nproc_per_node=$GPUS --master_port=$PORT ./tools/train.py $CONFIG --loss_reg $SEARCH_REG_LOSS --launcher pytorch;

# Train with default classification loss and default regression loss
python -m torch.distributed.launch --nproc_per_node=$GPUS --master_port=$PORT ./tools/train.py $CONFIG --launcher pytorch;

Acknowledgement

Thanks to MMDetection Team for their powerful deep learning detection framework. Thanks to Huawei Noah's Ark Lab AI Theory Group for their numerous V100 GPUs.

Citation

If you use this toolbox or benchmark in your research, please cite this project.

@inproceedings{
  liu2021loss,
  title={Loss Function Discovery for Object Detection via Convergence-Simulation Driven Search},
  author={Peidong Liu and Gengwei Zhang and Bochao Wang and Hang Xu and Xiaodan Liang and Yong Jiang and Zhenguo Li},
  booktitle={International Conference on Learning Representations},
  year={2021},
  url={https://openreview.net/forum?id=5jzlpHvvRk}
}
@article{mmdetection,
  title   = {{MMDetection}: Open MMLab Detection Toolbox and Benchmark},
  author  = {Chen, Kai and Wang, Jiaqi and Pang, Jiangmiao and Cao, Yuhang and
             Xiong, Yu and Li, Xiaoxiao and Sun, Shuyang and Feng, Wansen and
             Liu, Ziwei and Xu, Jiarui and Zhang, Zheng and Cheng, Dazhi and
             Zhu, Chenchen and Cheng, Tianheng and Zhao, Qijie and Li, Buyu and
             Lu, Xin and Zhu, Rui and Wu, Yue and Dai, Jifeng and Wang, Jingdong
             and Shi, Jianping and Ouyang, Wanli and Loy, Chen Change and Lin, Dahua},
  journal= {arXiv preprint arXiv:1906.07155},
  year={2019}
}
Owner
Peidong Liu(刘沛东)
Master Student in CS @ Tsinghua University. My research interest lies in scene understanding, visual tracking and AutoML for loss function.
Peidong Liu(刘沛东)
LAnguage Model Analysis

LAMA: LAnguage Model Analysis LAMA is a probe for analyzing the factual and commonsense knowledge contained in pretrained language models. The dataset

Meta Research 960 Jan 08, 2023
Source code for "Interactive All-Hex Meshing via Cuboid Decomposition [SIGGRAPH Asia 2021]".

Interactive All-Hex Meshing via Cuboid Decomposition Video demonstration This repository contains an interactive software to the PolyCube-based hex-me

Lingxiao Li 131 Dec 05, 2022
Translation-equivariant Image Quantizer for Bi-directional Image-Text Generation

Translation-equivariant Image Quantizer for Bi-directional Image-Text Generation Woncheol Shin1, Gyubok Lee1, Jiyoung Lee1, Joonseok Lee2,3, Edward Ch

Woncheol Shin 7 Sep 26, 2022
😮The official implementation of "CoNeRF: Controllable Neural Radiance Fields" 😮

CoNeRF: Controllable Neural Radiance Fields This is the official implementation for "CoNeRF: Controllable Neural Radiance Fields" Project Page Paper V

Kacper Kania 61 Dec 24, 2022
The repository includes the code for training cell counting applications. (Keras + Tensorflow)

cell_counting_v2 The repository includes the code for training cell counting applications. (Keras + Tensorflow) Dataset can be downloaded here : http:

Weidi 113 Oct 06, 2022
This is implementation of AlexNet(2012) with 3D Convolution on TensorFlow (AlexNet 3D).

AlexNet_3dConv TensorFlow implementation of AlexNet(2012) by Alex Krizhevsky, with 3D convolutiional layers. 3D AlexNet Network with a standart AlexNe

Denis Timonin 41 Jan 16, 2022
[ICSE2020] MemLock: Memory Usage Guided Fuzzing

MemLock: Memory Usage Guided Fuzzing This repository provides the tool and the evaluation subjects for the paper "MemLock: Memory Usage Guided Fuzzing

Cheng Wen 54 Jan 07, 2023
A facial recognition doorbell system using a Raspberry Pi

Facial Recognition Doorbell This project expands on the person-detecting doorbell system to allow it to identify faces, and announce names accordingly

rydercalmdown 22 Apr 15, 2022
YOLOX is a high-performance anchor-free YOLO, exceeding yolov3~v5 with ONNX, TensorRT, ncnn, and OpenVINO supported.

Introduction YOLOX is an anchor-free version of YOLO, with a simpler design but better performance! It aims to bridge the gap between research and ind

7.7k Jan 03, 2023
CRISCE: Automatically Generating Critical Driving Scenarios From Car Accident Sketches

CRISCE: Automatically Generating Critical Driving Scenarios From Car Accident Sketches This document describes how to install and use CRISCE (CRItical

Chair of Software Engineering II, Uni Passau 2 Feb 09, 2022
The modify PyTorch version of Siam-trackers which are speed-up by TensorRT.

SiamTracker-with-TensorRT The modify PyTorch version of Siam-trackers which are speed-up by TensorRT or ONNX. [Updating...] Examples demonstrating how

9 Dec 13, 2022
The code of "Dependency Learning for Legal Judgment Prediction with a Unified Text-to-Text Transformer".

Code data_preprocess.py: preprocess data for Dependent-T5. parameters.py: define parameters of Dependent-T5. train_tools.py: traning and evaluation co

1 Apr 21, 2022
PenguinSpeciesPredictionML - Basic model to predict Penguin species based on beak size and sex.

Penguin Species Prediction (ML) 🐧 👨🏽‍💻 What? 💻 This project is a basic model using sklearn methods to predict Penguin species based on beak size

Tucker Paron 0 Jan 08, 2022
Framework that uses artificial intelligence applied to mathematical models to make predictions

LiconIA Framework that uses artificial intelligence applied to mathematical models to make predictions Interface Overview Table of contents [TOC] 1 Ar

4 Jun 20, 2021
Computer Vision and Pattern Recognition, NUS CS4243, 2022

CS4243_2022 Computer Vision and Pattern Recognition, NUS CS4243, 2022 Cloud Machine #1 : Google Colab (Free GPU) Follow this Notebook installation : h

Xavier Bresson 142 Dec 15, 2022
PyTorch code of my ICDAR 2021 paper Vision Transformer for Fast and Efficient Scene Text Recognition (ViTSTR)

Vision Transformer for Fast and Efficient Scene Text Recognition (ICDAR 2021) ViTSTR is a simple single-stage model that uses a pre-trained Vision Tra

Rowel Atienza 198 Dec 27, 2022
To model the probability of a soccer coach leave his/her team during Campeonato Brasileiro for 10 chosen teams and considering years 2018, 2019 and 2020.

To model the probability of a soccer coach leave his/her team during Campeonato Brasileiro for 10 chosen teams and considering years 2018, 2019 and 2020.

Larissa Sayuri Futino Castro dos Santos 1 Jan 20, 2022
Code associated with the paper "Towards Understanding the Data Dependency of Mixup-style Training".

Mixup-Data-Dependency Code associated with the paper "Towards Understanding the Data Dependency of Mixup-style Training". Running Alternating Line Exp

Muthu Chidambaram 0 Nov 11, 2021
Natural Intelligence is still a pretty good idea.

Human Learn Machine Learning models should play by the rules, literally. Project Goal Back in the old days, it was common to write rule-based systems.

vincent d warmerdam 641 Dec 26, 2022
A program to recognize fruits on pictures or videos using yolov5

Yolov5 Fruits Detector Requirements Either Linux or Windows. We recommend Linux for better performance. Python 3.6+ and PyTorch 1.7+. Installation To

Fateme Zamanian 30 Jan 06, 2023