Jetson emulator object detection output is all gray

Hi, I am a newbie working with jetson nano. I found a simple code about streaming and object detection using the jetson emulator. I attached the code below and its output. The problem is the output which is all gray, but the classification task seems working.
how can I fix this?
Thanks in advance.

import jetson_emulator.inference as inference
import jetson_emulator.utils as utils

network = "ssd-mobilenet-v2"
net = inference.detectNet(network, threshold=0.5)
input_URI = "rtsp://jetson_emulator:554/detectNet/road_cam/4k" 
input  = utils.videoSource(input_URI, argv="")
output = utils.videoOutput("display://1", argv="")
img = input.Capture()
detections = net.Detect(img, "box")
output.SetStatus("{:s} | Network {:.0f} FPS".format(network, net.GetNetworkFPS()))
output.Render(img)
print("detected {:d} objects in image\n".format(len(detections)) )
for detection in detections:
	print("class_desc:", net.GetClassDesc(detection.ClassID))  
	print(detection)

Hi @user162134, if you have a Jetson Nano, try using the actual jetson-inference project instead:

The jetson_emulator is a third-party project for PC and non-GPU platforms that has a similar interface to jetson-inference. I would ask any questions about that on the jetson_emulator GitHub: GitHub - teavuihuang/jetson-emulator: The Jetson Emulator emulates the NVIDIA Jetson AI-Computer's Inference and Utilities API for image classification, object detection and image segmentation (i.e. imageNet, detectNet and segNet). The intended users are makers, learners, developers and students who are curious about AI computers and AI edge-computing but are not ready or able to invest in an actual device such as the NVIDIA Jetson Nano Developer Kit (https://developer.nvidia.com/embedded/jetson-nano-developer-kit). E.g. this allows every student in a computer class to have their own personal AI computer to explore and experiment with. This Jetson Emulator presents a pre-configured, ready-to-run kit with 2 virtual HDMI displays and 4 virtual live-cameras. This enables usage familiarisation with the Jetson API and experimentation with AI computer vision inference. It is a great way to quickly and easily get 'hands-on' with Jetson and experience the power of AI.

Hi @dusty_nv
Thank you for your reply,
I don’t have the real board now,
I have to prepare a code and if the test were ok, try it on the real board.
Do you mean I can’t do this test with the emulator and I have to buy the board?

I see - the emulator is meant for if you don’t have a real board. However the emulator is a third-party project that I don’t personally maintain or know, so you will want to file a question on their GitHub for jetson-emulator if you have issues with it.

If you have a real board, the actual jetson-inference project is for the real boards. It will work on the Nano because I test it on the real hardware.

This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.