Search Results for author:
Found 24 papers, 15 papers with code
Independent Component Alignment for Multi-Task Learning
In this work, we propose using a condition number of a linear system of gradients as a stability criterion of an MTL optimization.
Contour-based Interactive Segmentation
Recent advances in interactive segmentation (IS) allow speeding up and simplifying image editing and labeling greatly.
TR3D: Towards Real-Time Indoor 3D Object Detection
Our model with early feature fusion, which we refer to as TR3D+FF, outperforms existing 3D object detection approaches on the SUN RGB-D dataset.
Ranked #2 on
3D Object Detection
on S3DIS
(using extra training data)
Top-Down Beats Bottom-Up in 3D Instance Segmentation
Most 3D instance segmentation methods exploit a bottom-up strategy, typically including resource-exhaustive post-processing.
Ranked #3 on
3D Instance Segmentation
on S3DIS
(using extra training data)
Floorplan-Aware Camera Poses Refinement
Processing large indoor scenes is a challenging task, as scan registration and camera trajectory estimation methods accumulate errors across time.
Single-Stage 3D Geometry-Preserving Depth Estimation Model Training on Dataset Mixtures With Uncalibrated Stereo Data
On the contrary, we propose GP2, General-Purpose and Geometry-Preserving training scheme, and show that conventional SVDE models can learn correct shifts themselves without any post-processing, benefiting from using stereo data even in the geometry-preserving setting.
FCAF3D: Fully Convolutional Anchor-Free 3D Object Detection
Existing 3D object detection methods make prior assumptions on the geometry of objects, and we argue that it limits their generalization ability.
Ranked #5 on
3D Object Detection
on S3DIS
ImVoxelNet: Image to Voxels Projection for Monocular and Multi-View General-Purpose 3D Object Detection
To address this problem, we propose ImVoxelNet, a novel fully convolutional method of 3D object detection based on monocular or multi-view RGB images.
Ranked #1 on
Monocular 3D Object Detection
on SUN RGB-D
Reviving Iterative Training with Mask Guidance for Interactive Segmentation
We find that the models trained on a combination of COCO and LVIS with diverse and high-quality annotations show performance superior to all existing models.
Ranked #5 on
Interactive Segmentation
on DAVIS
Road images augmentation with synthetic traffic signs using neural networks
Such training data is obtained by embedding synthetic images of signs in the real photos.
Towards General Purpose Geometry-Preserving Single-View Depth Estimation
Our work shows that a model trained on this data along with conventional datasets can gain accuracy while predicting correct scene geometry.
Learning High-Resolution Domain-Specific Representations with a GAN Generator
Based on this finding, we propose LayerMatch scheme for approximating the representation of a GAN generator that can be used for unsupervised domain-specific pretraining.
Semi-Supervised Semantic Segmentation
Vocal Bursts Intensity Prediction
Foreground-aware Semantic Representations for Image Harmonization
Image harmonization is an important step in photo editing to achieve visual consistency in composite images by adjusting the appearances of foreground to make it compatible with background.
Decoder Modulation for Indoor Depth Completion
Our second contribution is a novel training strategy that allows us to train on a semi-dense sensor data when the ground truth depth map is not available.
Ranked #1 on
Depth Completion
on Matterport3D
IterDet: Iterative Scheme for Object Detection in Crowded Environments
Deep learning-based detectors usually produce a redundant set of object bounding boxes including many duplicate detections of the same object.
Ranked #1 on
Object Detection
on WiderPerson
f-BRS: Rethinking Backpropagating Refinement for Interactive Segmentation
We propose f-BRS (feature backpropagating refinement scheme) that solves an optimization problem with respect to auxiliary variables instead of the network inputs, and requires running forward and backward pass just for a small part of a network.
Ranked #8 on
Interactive Segmentation
on SBD
Training Deep SLAM on Single Frames
We train visual odometry model on synthetic data and do not use ground truth poses hence this model can be considered unsupervised.
Measuring robustness of Visual SLAM
We find that while in many cases the accuracy of SLAM is very good, the robustness is still an issue.
DISCOMAN: Dataset of Indoor SCenes for Odometry, Mapping And Navigation
We present a novel dataset for training and benchmarking semantic SLAM methods.
AdaptIS: Adaptive Instance Selection Network
Given an input image and a point $(x, y)$, it generates a mask for the object located at $(x, y)$.
Ranked #8 on
Panoptic Segmentation
on Mapillary val
Perceptual Image Anomaly Detection
We present a novel method for image anomaly detection, where algorithms that use samples drawn from some distribution of "normal" data, aim to detect out-of-distribution (abnormal) samples.
Scene Motion Decomposition for Learnable Visual Odometry
Optical Flow (OF) and depth are commonly used for visual odometry since they provide sufficient information about camera ego-motion in a rigid scene.
Double Refinement Network for Efficient Indoor Monocular Depth Estimation
Monocular depth estimation is the task of obtaining a measure of distance for each pixel using a single image.
Indoor Monocular Depth Estimation
Monocular Depth Estimation
Pose-based Deep Gait Recognition
In this paper, we present a new pose-based convolutional neural network model for gait recognition.
