Mifeng Technology has forged a collaborative partnership with Northwestern Polytechnical University and Xi'an Jiaotong University to unveil MEgoVista—a cutting-edge motion reconstruction framework designed from a first-person perspective. This innovative framework excels at precisely converting first-person videos into 4D data, catering specifically to the pre-training requirements of embodied models. The relevant research papers have already been published. At present, while first-person videos facilitate real-world data collection, the process of transforming them into 4D data presents numerous hurdles. These include challenges in striking a balance between scene coverage and reconstruction precision, inherent biases in distance estimation from a single camera, and tracking disruptions caused by occlusions during interactions involving multiple individuals.
