Split Neural Networks for Mobile Devices

Ushakov, Yury A.; Polezhaev, Petr; Shukhman, Aleksandr E.; Ushakova, Margarita; Nadezhda, M.V.

doi:10.1109/telfor.2018.8612133

Cited by 5 publications

(2 citation statements)

References 3 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Smile is an attempt to create a standard JSON [18] binary representation. Smile development started on 2010 led by Tatu Saloranta 136 , founder of FasterXML 137 while also being a Principal Software Engineer at Salesforce. Smile is released under the 2-clause BSD license 138 .…”

Section: Characteristicsmentioning

confidence: 99%

A Survey of JSON-compatible Binary Serialization Specifications

Viotti¹,

Kinderkhedia²

2022

Preprint

View full text Add to dashboard Cite

In this paper, we present the recent advances that highlight the characteristics of JSON-compatible binary serialization specifications. We motivate the discussion by covering the history and evolution of binary serialization specifications across the years starting from 1960s to 2000s and onwards. We analyze the use cases of the most popular serialization specifications across the industries. Drawing on the schema-driven (ASN.1, Apache Avro, Microsoft Bond, Cap'n Proto, FlatBuffers, Protocol Buffers, and Apache Thrift) and schema-less (BSON, CBOR, FlexBuffers, MessagePack, Smile, and UBJSON) JSON-compatible binary serialization specifications, we compare and contrast their inner workings through our analysis. We explore a set of non-standardized binary integer encoding techniques (ZigZag integer encoding and Little Endian Base 128 variable-length integer encoding) that are essential to understand the various JSON-compatible binary serialization specifications. We systematically discuss the history, the characteristics, and the serialization processes of the selection of schema-driven and schema-less binary serialization specifications and we identify the challenges associated with schema evolution in the context of binary serialization. Through reflective exercise, we explain our observations of the selection of JSON-compatible binary serialization specifications. This paper aims to guide the reader to make informed decisions on the choice between schema-driven or schema-less JSON-compatible binary serialization specifications.

show abstract

Section: Characteristicsmentioning

confidence: 99%

A Survey of JSON-compatible Binary Serialization Specifications

Viotti¹,

Kinderkhedia²

2022

Preprint

View full text Add to dashboard Cite

show abstract

“…To reduce the size and complexity of our model to operate on a smartphone, we froze our 86 MB 30,000-step TensorFlow model and converted our 23 MB frozen inference graph into a 22 MB TFLite model, which is an offline model optimized for smartphone devices requiring low latency and a small binary size (Ushakov et al, 2018). Table 1 shows experimental results of our proposed system for ten object categories carried out in five rounds of experiments.…”

Section: Tensorflowlite (Tflite)mentioning

confidence: 99%

Indoor 3d Interactive Asset Detection Using a Smartphone

Kostoeva

Rishi

Sapar

et al. 2019

Int. Arch. Photogramm. Remote Sens. Spatial Inf. Sci.

View full text Add to dashboard Cite

<p><strong>Abstract.</strong> Building floor plans with locations of safety, security and energy assets such as IoT sensors, thermostats, fire sprinklers, EXIT signs, fire alarms, smoke detectors, routers etc. are vital for climate control, emergency security, safety, and maintenance of building infrastructure. Existing approaches to building survey are manual, and usually involve an operator with a clipboard and pen, or a tablet enumerating and localizing assets in each room. In this paper, we propose an interactive method for a human operator to use an app on a smart phone to (a) create the 2D layout of a room, (b) detect assets of interest, and (c) localize them within the layout. We use deep learning methods to train a neural network to recognize assets of interest, and use human in the loop interactive methods to correct erroneous recognitions by the networks. These corrections are then used to improve the accuracy of the system over time as the inspector moves from one room to another in a given building or from one building to the next; this progressive training and testing mechanism makes our system useful in building inspection scenarios where a given class of assets in a building are same instantiation of that object category, thus reducing the problem to instance, rather than category recognition. Experiments show our proposed method to achieve accuracy rate of 76% for testing 102 objects across 10 classes.</p>

show abstract