文章目录
- Maintaining Performance with Less Data
- DINO: DETR with Improved DeNoising Anchor Boxes for End-to-End Object Detection(ICLR2023补)
- S^3 FD: Single Shot Scale-invariant Face Detector(2017)
- Object Detectors Emerge in Deep Scene CNNs(2015)
- Understanding the Effective Receptive Field in Deep Convolutional Neural Networks(2017)
- GooGleNet/Inception(2014)
- Distilling the Knowledge in a Neural Network(2015)(d补)
- UNet(2015)
- ResNet(CVPR2016)
- DenseNet(2017)
- ResNeXt(2017)
- SENet(2018)
- CSPNet(2019)
- EfficientNet(2020)
- SqueezeNet(2016)
- MobileNets(2017)
- ShuffleNet(2017)
- MobileNetV2(2019)
- ShuffleNet V2(2018)
- PeleeNet(2019)
- MnasNet(2019)
- MobileNetsV3(2019)
- Once-For-All(OFA2020)
- FPN(CVPR2017)
- PAN(2018.9)
- PanNet: A deep network architecture for pan-sharpening(ICCV 2017)
- Benchmarking Detection Transfer Learning with Vision Transformers(2021.11)
- Exploring Plain Vision Transformer Backbones for Object Detection(2022.6)
- Masked Autoencoders Are Scalable Vision Learners(2021.12)
- Masked Autoencoders As Spatiotemporal Learners(2022.10)
- Scaling Language-Image Pre-training via Masking(2023.5)
- Multi-scale structural Kernel Representation for Object Detection(2018.4)
- RetinaNet(2018)
- DETR(2020)
- One Thumbnail Image Contains All You Need for Recognition(2020.10)
- One thumbnail image is enough for object detection(2022.12)
- Distilling Object Detectors via Decoupled Features(CVPR2021)
- Swin Transformer: Hierarchical Vision Transformer using Shifted Windows(ICCV2021)
- Swin Transformer V2 Scaling Up Capacity and Resolution(CVPR2022)
- CSWin Transformer: A General Vision Transformer Backbone with Cross-Shaped Windows(CVPR2022)
- UNIPOSE: DETECTING ANY KEYPOINTS(2023.10.12)
- HUMAN TOMATO: T EXT - ALIGNED WHOLE-BODY MOTION GENERATION(2023.10.19)
- Run, Don‘t Walk: Chasing Higher FLOPS for Faster Neural Networks(CVPR2023)
Maintaining Performance with Less Data
很好
提出了一种用于图像分类的神经网络训练的新方法,动态地减少输入数据
描述了三种动态引入数据到模型的新方法,每种方法都减少了执行的评估量:
– Data Step
– Data Increment
– Data Cut
Model如下,key:加了数据选择阶段
数据收集–>数据选择–>数据增强–>训练–>数据选择–>数据增强–>训练…–>训练完成–>输出
DINO: DETR with Improved DeNoising Anchor Boxes for End-to-End Object Detection(ICLR2023补)
hh
提出了DINO(带改进去噪锚盒的DETR),一种最先进的端到端目标检测器。DINO通过使用对比方法进行去噪训练,使用混合查询选择方法进行