Description
机器人抓不住玻璃杯?不是手笨,是深度相机 “视障了”!透明、反光物体一出现,深度图就满是黑窟窿…
蚂蚁灵波开源的高精度空间感知模型LingBot-Depth,靠 “给深度图做完形填空” 的深度掩码建模,把带黑窟窿的原始深度相机深度图,脑补成轮廓清晰的完整深度图。不管是机器人抓取、3D 追踪还是 SLAM 重建,下游任务直接起飞~
官网:https://technology.robbyant.com/lingbot-depth
Github:https://github.com/robbyant/lingbot-depth
魔搭社区:https://www.modelscope.cn/collections/Robbyant/LingBot-Depth
HuggingFace: https://huggingface.co/robbyant/lingbot-depth
领英:https://www.linkedin.com/posts/robbyant_depth-sensors-struggle-with-glass-and-reflective-activity-7421610955719131136-bIMX