Proposes PointSetNet for single-image 3D reconstruction, using Earth Mover's Distance and multi-sample prediction, outperforming SOTA.

Haoqiang Fan, Hao Su, Leonidas Guibas

3D-EPN combines shape synthesis for high-accuracy 3D shape completion.

Angela Dai, Charles Ruizhongtai Qi, Matthias Nießner

Proposes shallow-wide residual networks outperforming ResNet-200, enhancing image classification and segmentation with improved efficiency.

Zifeng Wu, Chunhua Shen, Anton van den Hengel

Proposes a DenseNet-based fully convolutional network for semantic segmentation, achieving state-of-the-art IoU 66.9% on CamVid with only 9.4M parameters.

Simon Jégou, Michal Drozdzal, David Vazquez et al.

MV3D network fuses LIDAR and RGB data, achieving a 30% improvement in 3D detection accuracy on the KITTI dataset.

Xiaozhi Chen, Huimin Ma, Ji Wan et al.

iCaRL combines nearest-mean-of-exemplars classification with rehearsal and distillation, enabling long-term incremental learning and mitigating catastrophic forgetting.

Sylvestre-Alvise Rebuffi, Alexander Kolesnikov, Georg Sperl et al.

Proposes Deep Feature Flow, running expensive CNNs only on sparse key frames and propagating features via optical flow, achieving 10x speedup.

Xizhou Zhu, Yuwen Xiong, Jifeng Dai et al.

Proposes conditional GAN with U-Net and PatchGAN for image translation, achieving state-of-the-art results with high detail fidelity.

Phillip Isola, Jun-Yan Zhu, Tinghui Zhou et al.

Integrating deep CNNs (YOLOv4) with GMM background modeling and multisensor fusion achieves 85% detection accuracy and 92% tracking continuity in maritime videos.

D. K. Prasad, D. Rajan, L. Rachmawati et al.

DSAC improves camera localization accuracy by 7.3% through differentiable RANSAC.

Eric Brachmann, Alexander Krull, Sebastian Nowozin et al.

ResNeXt introduces the 'cardinality' dimension, outperforming ResNet by increasing transformation paths, achieving higher accuracy with similar complexity.

Saining Xie, Ross Girshick, Piotr Dollár et al.

Constructed the COMICS dataset with 1.2 million panels; proposed three closure tasks; deep models lag behind humans, highlighting challenges in multimodal narrative understanding.

Mohit Iyyer, Varun Manjunatha, Anupam Guha et al.

End-to-end deep architecture based on R-MAC with triplet loss significantly improves image retrieval accuracy, achieving over 94% MAP on standard benchmarks.

Albert Gordo, Jon Almazan, Jerome Revaud et al.

3D-GAN generates high-quality 3D objects, enhancing recognition performance.

Jiajun Wu, Chengkai Zhang, Tianfan Xue et al.

Proposes Xception, a CNN based on depthwise separable convolutions, outperforming Inception V3 with similar parameters.

François Chollet

Proposed SRGAN employs GAN with perceptual loss for 4× super-resolution, significantly enhancing perceptual quality.

Christian Ledig, Lucas Theis, Ferenc Huszar et al.

Proposes a novel unsupervised monocular depth estimation method with left-right consistency, outperforming supervised methods on KITTI.

Clément Godard, Oisin Mac Aodha, Gabriel J. Brostow

Uses GAN-based latent space modeling with constrained optimization for realistic, user-controlled image editing, achieving near real-time performance.

Jun-Yan Zhu, Philipp Krähenbühl, Eli Shechtman et al.

Deep learning framework combining RNN and CNN decodes EEG signals to classify 40 ImageNet categories with 83% accuracy, enabling brain-driven image recognition.

Concetto Spampinato, Simone Palazzo, Isaak Kavasidis et al.

Proposes filter pruning based on `1-norm` importance, reducing FLOPs by up to 34% for VGG-16 and 38% for ResNet-110 on CIFAR-10, with minimal accuracy loss.

Hao Li, Asim Kadav, Igor Durdanovic et al.