NurtureNet: A Multi-task Video-based Approach for Newborn Anthropometry
Authors:
Yash Khandelwal,
Mayur Arvind,
Sriram Kumar,
Ashish Gupta,
Sachin Kumar Danisetty,
Piyush Bagad,
Anish Madan,
Mayank Lunayach,
Aditya Annavajjala,
Abhishek Maiti,
Sansiddh Jain,
Aman Dalmia,
Namrata Deka,
Jerome White,
Jigar Doshi,
Angjoo Kanazawa,
Rahul Panicker,
Alpan Raval,
Srinivas Rana,
Makarand Tapaswi
Abstract:
Malnutrition among newborns is a top public health concern in develo** countries. Identification and subsequent growth monitoring are key to successful interventions. However, this is challenging in rural communities where health systems tend to be inaccessible and under-equipped, with poor adherence to protocol. Our goal is to equip health workers and public health systems with a solution for c…
▽ More
Malnutrition among newborns is a top public health concern in develo** countries. Identification and subsequent growth monitoring are key to successful interventions. However, this is challenging in rural communities where health systems tend to be inaccessible and under-equipped, with poor adherence to protocol. Our goal is to equip health workers and public health systems with a solution for contactless newborn anthropometry in the community.
We propose NurtureNet, a multi-task model that fuses visual information (a video taken with a low-cost smartphone) with tabular inputs to regress multiple anthropometry estimates including weight, length, head circumference, and chest circumference. We show that visual proxy tasks of segmentation and keypoint prediction further improve performance. We establish the efficacy of the model through several experiments and achieve a relative error of 3.9% and mean absolute error of 114.3 g for weight estimation. Model compression to 15 MB also allows offline deployment to low-cost smartphones.
△ Less
Submitted 8 May, 2024;
originally announced May 2024.
Pathways towards 30% efficient single-junction perovskite solar cells and the role of mobile ions
Authors:
Jonas Diekmann,
Pietro Caprioglio,
Moritz H. Futscher,
Vincent M. Le Corre,
Sebastian Reichert,
Frank Jaiser,
Malavika Arvind,
Lorena Perdigon Toro,
Emilio Gutierrez-Partida,
Francisco Pena-Camargo,
Carsten Deibel,
Bruno Ehrler,
Thomas Unold,
Thomas Kirchartz,
Dieter Neher,
Martin Stolterfoht
Abstract:
Perovskite semiconductors have demonstrated outstanding external luminescence quantum yields, enabling high power conversion efficiencies (PCE). However, the precise conditions to advance to an efficiency regime above monocrystalline silicon cells are not well understood. Here, we establish a simulation model that well describes efficient p-i-n type perovskite solar cells and a range of different…
▽ More
Perovskite semiconductors have demonstrated outstanding external luminescence quantum yields, enabling high power conversion efficiencies (PCE). However, the precise conditions to advance to an efficiency regime above monocrystalline silicon cells are not well understood. Here, we establish a simulation model that well describes efficient p-i-n type perovskite solar cells and a range of different experiments. We then study important device and material parameters and we find that an efficiency regime of 30% can be unlocked by optimizing the built-in potential across the perovskite layer by using either highly doped (10^19 cm-3), thick transport layers (TLs) or ultrathin undoped TLs, e.g. self-assembled monolayers. Importantly, we only consider parameters that have been already demonstrated in recent literature, that is a bulk lifetime of 10 us, interfacial recombination velocities of 10 cm/s, a perovskite bandgap of 1.5 eV and an EQE of 95%. A maximum efficiency of 31% is predicted for a bandgap of 1.4 eV. Finally, we demonstrate that the relatively high mobile ion density does not represent a significant barrier to reach this efficiency regime. Thus, the results of this paper promise continuous PCE improvements until perovskites may become the most efficient single-junction solar cell technology in the near future.
△ Less
Submitted 25 April, 2021; v1 submitted 16 October, 2019;
originally announced October 2019.
