Posters
Posters that caught my eyes in the 6 Poster sessions from Tuesday to Thursday.
Theory
- Grid R-CNN (SenseTime) Note
- Class-Balanced Loss Based on Effective Number of Samples
- Variational Autoencoders Pursue PCA Directions (by Accident)
- Pixel-Adaptive Convolutional Neural Networks
- Explainability Methods for Graph Convolutional Neural Networks
- Building Detail-Sensitive Semantic Segmentation Networks with Polynomial Pooling
- SparseFool: a few pixels make a big difference
- Learning Multi-Class Segmentations From Single-Class Datasets
- Effective Aesthetics Prediction with Multi-level Spatially Pooled Features
- Learning Not to Learn: Training Deep Neural Networks with Biased Data
- Joint Manifold Diffusion for Combining Predictions on Decoupled Observations
- Kernel Transformer Networks for Compact Spherical Convolution (check out the non-euclidean bbox!)
depth estimation
- Neural RGB->D Sensing: Depth and Uncertainty from a Video Camera
- Monocular 3D Object Detection Leveraging Accurate Proposals and Shape Reconstruction
- CAM-Convs: Camera-Aware Multi-Scale Convolutions for Single-View Depth
Model compression
- ChamNet: Towards Efficient Network Design through Platform-Aware Model Adaptation
- Accelerating Convolutional Neural Networks via Activation Map Compression
- Structured Pruning of Neural Networks with Budget-Aware Regularization
- Partial Order Pruning: for Best Speed/Accuracy Trade-off in Neural Architecture Search (check out the very accurate latency LUT!)
- ESPNetv2: A Light-weight, Power Efficient, and General Purpose Convolutional Neural Network
Autonomous driving
- End-to-end Interpretable Neural Motion Planner
- Selective Sensor Fusion for Neural Visual-Inertial Odometry (Oxford)
- A Parametric Top-View Representation of Complex Road Scenes
- Monocular Depth Estimation Using Relative Depth Maps (indoor scenes)
- PointNetLK: Robust & Efficient Point Cloud Registration using PointNet
- Adaptive NMS: Refining Pedestrian Detection in a Crowd
- http://cvlab.cse.msu.edu/pdfs/Brazil_Liu_CVPR2019.pdf
- Deep Incremental Hashing Network for Efficient Image Retrieval
- Learning for Single-Shot Confidence Calibration in Deep Neural Networks through Stochastic Inferences
- Rules of the Road: Predicting Driving Behavior with a Convolutional Model of Semantic Interactions
- Siamese Cascaded Region Proposal Networks for Real-Time Visual Tracking
- Triangulation Learning Network: from Monocular to Stereo 3D Object Detection
- Leveraging Crowdsourced GPS Data for Road Extraction from Aerial Imagery
- Multi-Task Multi-Sensor Fusion for 3D Object Detection
- Rare Event Detection using Disentangled Representation Learning
video detection/segmentation
- FEELVOS: Fast End-to-End Embedding Learning for Video Object Segmentation
- DFANet: Deep Feature Aggregation for Real-Time Semantic Segmentation
TODO
posters
- noise aware unsupervised deep lidar stereo fusion
- an attention based recurrent neural network for vehicle tailgate recognition
- superdepth: monocular
- ROI-10D: Monocular Lifting of 2d detection to 6d pose and metric shape
- TASCNet Learning to fuse things and stuff
- PackNet SfM
- SPIGAN previleged adversarial learning from simulation