occlusion-handling

Deep Occlusion-Aware Instance Segmentation with Overlapping BiLayers [CVPR 2021]

Behind the Curtain: Learning Occluded Shapes for 3D Object Detection

This is the official Python and C++ implementation repository for a paper entitled "Track Initialization and Re-Identification for 3D Multi-View Multi-Object Tracking", Information Fusion (http://arxiv.org/abs/2405.18606).

Master thesis project realized in partnership with Aalto University and Finnish Geospatial Research Institute of Finland. Stereo Camera-LiDAR calibration. Covers: mono and stereo camera calibration, Camera-LIDAR synchronisation, camera-based and LiDAR-based 3D reconstruction, Camera-LiDAR extrinsic calibration, Occlusion handling.

An instance segmentation challenge on Basketball images, with a particular focus on occlusion resolution. An opportunity to publish at MMSports @ ACMMM and to win $1000.

This is the official Python and C++ implementation repository for a paper entitled "Visual multi-object tracking with re-identification and occlusion handling using labeled random finite sets", Pattern Recognition (https://arxiv.org/pdf/2407.08872).

Noise Aware OSTrack

Public toolkit for HOOT: Heavy Occlusions for Object Tracking Benchmark.

🏴‍☠️🏴‍☠️Tells a robber in a balaclava from a customer in a surgical mask — real-time concealed-face detection for retail CCTV. Pose-based orientation, occlusion-hardened tracking, zero training.🏴‍☠️

3D normal map estimation pipeline using low resolution and highly occluded images

Game-theoretic occlusion-aware planner deployed on a full-scale autonomous vehicle at IIT Madras

This face recognition implementation is capable of recognizing faces with a certain level of occlusion, this includes faces wearing masks. You can also add new users manually by adding a photo in the images folder.

Build and explore GPT-5.4-mini stories with embedded ASCII art using this lightweight HTML narrative engine.

Unsupervised deep framework that animates a source image using motion extracted from a driving video through learned keypoints, local affine transforms, dense motion fields, and occlusion-aware generation.

Temporal 3D semantic occupancy grid for autonomous driving. Fuses sparse LiDAR depth and camera segmentation into a voxel map, then uses 5-frame ego-motion memory to keep occluded vehicles visible. Validated on 2,000 CARLA frames (+7.95 pp vehicle IoU in high-disagreement scenes).

Face Under Mask Reconstruction , Image Inpainting, and Object Detection and Segmentation with the Detectron2 library

Knowledge distillation for masked FER using ResNet-18 in PyTorch.

Comparative analysis of DeepOCSORT, StrongSORT, and ByteTrack for robust tracking under heavy occlusion.