Case studies and mentions

The Airbnb engineering and data science team applies machine learning using TensorFlow to classify images and detect objects at scale, helping to improve the guest experience.

ML helps with monitoring changes to the Earth's surface for urban planning, fighting illegal construction and mapping damage and landscape changes caused by natural catastrophes.

Arm NN for Android Neural Networks API (NNAPI) provides a Hardware Abstraction Layer (HAL) that targets Arm Mali GPUs and leads to more than a 4x performance boost to machine learning frameworks such as TensorFlow Lite.

Carousell builds machine learning models with deep image and natural language understanding using TensorFlow on Google Cloud ML. Sellers benefit from a simplified posting experience with image recognition, and buyers discover more relevant listings through recommendations and image search.

CEVA’s NeuPro and CEVA-XM AI processors for Deep Learning and AI inferencing at the edge automatically convert TensorFlow trained networks for use in real-time embedded devices using the CEVA CDNN Compiler.

China Mobile has created a deep learning system using TensorFlow that can automatically predict cutover time window, verify operation logs, and detect network anomalies. This has already successfully supported the world’s largest relocation of hundreds of millions IoT HSS numbers.

Advances in artificial intelligence and the maturity of TensorFlow enabled the Coca-Cola Company to finally achieve a long-sought frictionless proof-of-purchase capability.

Using Tensorflow, GE Healthcare is training a neural network to identify specific anatomy during brain magnetic resonance imaging (MRI) exams to help improve speed and reliability.

Google uses TensorFlow to power ML implementations in products like Search, Gmail, and Translate, to aid researchers in new discoveries, and even to forge advances in humanitarian and environmental challenges.

This work has resulted in up to 2.8x performance improvement which benefits the TensorFlow community and a wide range of customers using TensorFlow on Intel platforms

Kakao Mobility uses TensorFlow & TensorFlow Serving to predict the probability of trip completed rate when we dispatch drivers to fulfill ride hailing requests.

The Lenovo LiCO platform accelerates AI training and traditional High Performance Computing, and optimizes deep learning training with TensorFlow integration and optimization. LiCO provides various built-in TensorFlow models and supports optimized distributed training of these models.

The Liulishuo algorithm team first applied TensorFlow to its internal machine learning project in early 2016. This easy-to-use machine learning framework helped the team build an application to teach English.

Using TensorFlow NAVER Shopping automatically matches over 20 million newly registered products a day to around 5,000 categories in order to organize products systematically and allow easier searching for users.

NERSC and NVIDIA succeeded at scaling a scientific Deep Learning application to 27,000+ Nvidia V100 Tensor Core GPUs, breaking the ExaFLOP barrier in the process.

Using TensorFlow, deep transfer learning and generative modeling, PayPal has been able to recognize complex temporally varying fraud patterns to increase fraud decline accuracy while improving experience of legitimate users through increased precision in identification.

Qualcomm optimizes and accelerates TensorFlow and TensorFlow Lite models on Snapdragon mobile platforms, and across chipset portfolios designed for IoT, compute, XR and automotive.

Disease classification and segmentation were performed on retinal OCT images using TensorFlow. The three disease types were classified as either choroidal neovascularization, vitreous warts or diabetic retinal edema. After segmentation, Sinovation Ventures provided the boundary of the suspected lesions in the imaging.

Swisscom leverages TensorFlow's capacity to deeply customize machine learning models to classify text and determine the intent of their customers upon receiving their calls.

Twitter used TensorFlow to build their "Ranked Timeline," allowing users to ensure that they don't miss their most important tweets even if they follow thousands of users.

VSCO used TensorFlow Lite to develop the “For This Photo” feature, which uses on-device machine learning to identify what kind of photo someone is editing and then suggest relevant presets from a curated list.

WPS Office implements multiple business scenarios, such as on-device image recognition and image OCR based on TensorFlow.