Spiking-YOLO uses spiking neural networks for energy-efficient object detection, reducing energy consumption by 280 times compared to Tiny YOLO.

Seijoon Kim, Seongsik Park, Byunggook Na et al.

Group-wise correlation stereo network (GwcNet) constructs cost volumes via feature grouping, achieving state-of-the-art accuracy with fewer parameters.

Xiaoyang Guo, Kai Yang, Wukui Yang et al.

Introduced DDR network, achieving 5.9% improvement in 3D semantic scene completion with 79% fewer parameters.

Jie Li, Yu Liu, Dong Gong et al.

Proposes R-MVSNet with convolutional GRU for recursive cost volume regularization, reducing memory by over 4.6×, enabling high-res large-scale 3D reconstruction.

Yao Yao, Zixin Luo, Shiwei Li et al.

Proposed a self-supervised monocular endoscopy depth estimation method leveraging SfM sparse supervision, achieving sub-millimeter accuracy without manual labels.

Xingtong Liu, Ayushi Sinha, Masaru Ishii et al.

BigEarthNet, a large-scale multi-label Sentinel-2 dataset with 590,326 images, outperforms ImageNet pre-trained models in land-cover scene classification.

Gencer Sumbul, Marcela Charfuelan, Begüm Demir et al.

Using CNN nearest-neighbor retrieval, the authors build ciFAIR and show CIFAR accuracy is inflated by roughly 9–14% relatively.

Björn Barz, Joachim Denzler

Constructed FaceForensics++ dataset with 1.8M images; proposed CNN-based face forgery detector achieving >95% accuracy, outperforming humans.

Andreas Rössler, Davide Cozzolino, Luisa Verdoliva et al.

MONet combines end-to-end VAE and recurrent attention for unsupervised scene decomposition, identifying objects and background.

Christopher P. Burgess, Loic Matthey, Nicholas Watters et al.

Proposes Hybrid Task Cascade (HTC) framework, integrating detection and segmentation in multi-stage, achieving 48.6 Mask AP on MSCOCO, surpassing previous methods.

Kai Chen, Jiangmiao Pang, Jiaqi Wang et al.

DeepSDF learns continuous signed distance functions for high-quality 3D shape representation, interpolation, and completion.

Jeong Joon Park, Peter Florence, Julian Straub et al.

UPSNet achieves state-of-the-art panoptic segmentation via pixel-wise classification on Cityscapes.

Yuwen Xiong, Renjie Liao, Hengshuang Zhao et al.

Proposes Panoptic FPN, extending Mask R-CNN with shared FPN backbone and lightweight semantic branch, achieving top performance on instance and semantic segmentation.

Alexander Kirillov, Ross Girshick, Kaiming He et al.

Proposes pseudo-LiDAR, converting image depth maps into point clouds, boosting KITTI 3D detection from 22% to 74%.

Yan Wang, Wei-Lun Chao, Divyansh Garg et al.

CROSR enhances open-set recognition by combining classification and reconstruction learning, outperforming existing models.

Ryota Yoshihashi, Wen Shao, Rei Kawakami et al.

PointRCNN employs a two-stage framework, directly generating high-quality 3D proposals from raw point clouds, outperforming existing methods.

Shaoshuai Shi, Xiaogang Wang, Hongsheng Li

PartNet: a large-scale dataset with 573,585 fine-grained, hierarchical 3D parts, advancing shape understanding.

Kaichun Mo, Shilin Zhu, Angel X. Chang et al.

Proposes a Transformer-based video action recognition model achieving 25.0% mAP using only RGB frames.

Rohit Girdhar, João Carreira, Carl Doersch et al.

Proposed Scene Graph Auto-Encoder (SGAE) with shared dictionary D achieves 127.8 CIDEr-D on MS-COCO, enhancing reasoning in image captioning.

Xu Yang, Kaihua Tang, Hanwang Zhang et al.

Proposes attraction field-based region segmentation for line detection, achieving 4.5% performance gain with end-to-end deep learning.

Nan Xue, Song Bai, Fudong Wang et al.