Multimodal Instance Segmentation
Multimodal instance segmentation extends classical instance segmentation — which assigns a per-pixel mask and a class label to every individual object in an image — by incorporating complementary sensor streams such as depth maps, LiDAR point clouds, or infrared frames. Fusing these modalities helps the model handle ambiguous appearances, low light, and occlusion that trip up RGB-only systems.
Source record
Citations copied verbatim from the method’s source record. No claim-level verification is inferred from them.
- He, K., Gkioxari, G., Dollar, P., & Girshick, R. (2017). Mask R-CNN. Proceedings of the IEEE International Conference on Computer Vision (ICCV), 2961–2969. · DOI 10.1109/ICCV.2017.322
- Instance segmentation. Wikipedia. · URL
Curated claims
Claims persisted in the evidence ledger, each with its own assessment.
This view does not invent a claim assessment when the ledger has none.
Related methods
Generated from the method graph and shown as machine-suggested relations — no evidence claim is inferred.