Can Jetson Orin Nano handle four 720p60 network camera streams for object detection?

Issue Overview

The main question revolves around the capability of the Nvidia Jetson Orin Nano 8GB variant to handle four network camera streams for object detection. Specifically, the user is inquiring about:

• Supporting 4 network camera streams at 1280×720 resolution and 60 FPS
• Using a custom camera setup with Arducam B0348 2MP Global Shutter cameras capable of 120 FPS
• Streaming over ethernet network
• Receiving 60 FPS from each camera at the endpoint

The user is seeking a quick response to make an informed decision about placing an order for the Jetson Orin Nano.

Possible Causes

While there are no specific issues mentioned, potential concerns that might arise in this scenario include:

1. Hardware limitations: The Jetson Orin Nano might not have sufficient processing power or memory to handle multiple high-resolution streams simultaneously.

2. Decoder capabilities: The hardware decoder might not support the required number of streams or resolution.

3. Network bandwidth: Ethernet network might not provide sufficient bandwidth for multiple high-resolution streams.

4. Software compatibility: Object detection algorithms might not be optimized for the Jetson Orin Nano’s architecture.

5. Thermal constraints: Running multiple streams and performing object detection might lead to thermal issues.

Troubleshooting Steps, Solutions & Fixes

Based on the information provided in the forum discussion, here are the key points and solutions:

1. Hardware Decoder Capability:

   • The Jetson Orin Nano’s hardware decoder is capable of handling 4x 720p60 streams simultaneously.
   • This meets the user’s requirement of four 1280×720 resolution streams at 60 FPS.

2. Verifying Specifications:

   • Refer to the official Jetson Orin Nano datasheet for detailed specifications:

     https://developer.nvidia.com/downloads/assets/embedded/secure/jetson/orin_nano/docs/jetson_orin_nano_ds
     

   • Always check the most recent datasheet as specifications may be updated.

3. Hardware Encoder Limitation:

   • The Jetson Orin Nano does not have a hardware encoder.
   • If hardware encoding is required for your application, consider using the Orin NX or Xavier NX models instead.

4. Object Detection Performance:

   • While the hardware decoder can handle the streams, the actual performance of object detection will depend on:
     • The specific object detection algorithm used
     • The complexity of the scenes being analyzed
     • Any additional processing required by your application
   • Benchmark your specific use case on the Jetson Orin Nano to ensure it meets your performance requirements.

5. Network Configuration:

   • Ensure your network infrastructure can handle the bandwidth required for four 720p60 streams.
   • Use Cat6 or better Ethernet cables to support the high data throughput.
   • Consider using a managed switch with QoS (Quality of Service) features to prioritize video traffic.

6. Thermal Management:

   • Implement proper cooling solutions for the Jetson Orin Nano, especially when running at full capacity.
   • Monitor temperatures during extended use to ensure they remain within acceptable limits.

7. Software Optimization:

   • Utilize NVIDIA’s optimized libraries and frameworks for object detection, such as TensorRT, to maximize performance on the Jetson platform.
   • Consider using NVIDIA DeepStream SDK for efficient multi-stream processing and analytics.

8. Alternative Options:

   • If after testing you find the Jetson Orin Nano insufficient for your needs, consider:
     • Upgrading to the Jetson Orin NX or Xavier NX for additional performance and hardware encoding capabilities.
     • Distributing the workload across multiple Jetson devices if a single unit cannot meet the performance requirements.

By following these steps and considerations, you should be able to determine if the Jetson Orin Nano 8GB variant is suitable for your application involving four 720p60 network camera streams and object detection.