Welcome to Jetson Orin NANO Developer Kit repository
02.Person object detection
TAO PeopleNet Pedestrian Object Detection
TAO PeopleNet Pedestrian Object Detection1. Case Goal and Processing Flow2. Pre-Run Checks3. Starting Detection4. USB Camera Real-Time Detection4.1 Direct Run4.2 Common Parameters5. CSI Camera Real-Time Detection5.1 Direct Run5.2 Common Parameters6. Common Issues
Applicable environment: Jetson Orin Nano 8GB, JetPack 7.2.1, DeepStream 9.1
After setting the power mode to MAXN, running the case may pop up an over-current warning; just ignore it.
1. Case Goal and Processing Flow
PeopleNet detects classes such as person, face, and bag in video frames, and outputs classes, confidence, and rectangles. It does not output pixel-level human silhouettes, nor does it automatically provide cross-frame identity IDs.
video/USB/CSI → PeopleNet TensorRT inference → OSD detection boxes ├→ desktop preview └→ video mode MP4For human silhouettes use PeopleSemSegNet; for trajectory IDs add a Tracker; for cross-camera appearance matching use ReIdentificationNet.
2. Pre-Run Checks
Video mode runs in a Jetson desktop/VNC graphical session:
xxxxxxxxxxcd ~/DeepStream/tao-toolkit-shell-demols -lh video_in/Retail02_short.mp4ls -ld video_outConfirm the material exists and the output directory is writable. USB/CSI modes also require connecting the camera and confirming the device is not occupied by other programs.
3. Starting Detection
Run the preset video:
xxxxxxxxxx./run_video.sh peoplenetThe current video input is video_in/Retail02_short.mp4, and the output is video_out/out_peoplenet.mp4. Detection boxes are displayed frame by frame; the class text depends on the order of the labels file and class IDs.

After waiting for EOS, check:
xxxxxxxxxxgst-discoverer-1.0 video_out/out_peoplenet.mp4The input file is specified by the case configuration. You can put a new video into Jetson's video_in/ and replace it using the file name used in the configuration, or check the configuration and modify the input source. Keep the original material before replacing; use gst-discoverer-1.0 to check the new material's encoding and confirm the device can decode it. The output is fixed to video_out/out_peoplenet.mp4; rename old results first when you need to keep them.
4. USB Camera Real-Time Detection
4.1 Direct Run
Start from the Jetson desktop/VNC:
xxxxxxxxxxcd ~/DeepStream/tao-toolkit-shell-demo./run_usb.sh peoplenetUSB mode is a live preview and does not generate this case's MP4. The class, confidence, and box meanings in the frame are the same as in video mode; the actual input format is determined by the camera.

To stop, press Ctrl+C once in the launching terminal, wait for App run successful, and confirm the device is released. PeopleNet USB was verified on /dev/video0 YUYV 640x480@30; the specific performance does not represent other cameras.
4.2 Common Parameters
Before running, you can check the formats, resolutions, and frame rates supported by the camera:
xxxxxxxxxxv4l2-ctl -d /dev/video0 --list-formats-ext
Note: the USB pipeline of Object Embedding Vector Generation is fixed to YUY2, so among the parameters below, choose the YUY2 parameters
x# View all parameters./run_usb.sh peoplenet --help
# Use another video device./run_usb.sh peoplenet --device /dev/video1
# Choose from the configurations found above that the camera supports; here we use 1280×720, 9FPS./run_usb.sh peoplenet --width 1280 --height 720 --fps 95. CSI Camera Real-Time Detection
5.1 Direct Run
The course CSI entry supports live preview. Connect a compatible camera, confirm the driver and Argus service are working, then run the corresponding command in a Jetson desktop/VNC graphical terminal:
xxxxxxxxxxcd ~/DeepStream/tao-toolkit-shell-demo./run_csi.sh peoplenetCSI uses the Jetson Argus link; the existence of a USB device node does not prove CSI streaming works. Camera mode is for live preview and does not generate the video mode's MP4.

5.2 Common Parameters
Before running, you can check the formats, resolutions, and frame rates supported by the camera; this must be run in a graphical interface:
xxxxxxxxxxgst-launch-1.0 nvarguscamerasrc
xxxxxxxxxx# View all parameters./run_csi.sh peoplenet --help
# Use the second CSI camera./run_csi.sh peoplenet --sensor-id 1
# Adjust input size and frame rate./run_csi.sh peoplenet --width 1920 --height 1080 --fps 306. Common Issues
| Symptom | Check and Handling |
|---|---|
| No person boxes | Confirm the input frame has targets and the pipeline is running; occlusion, scale, and lighting all affect detection. |
| Classes displayed incorrectly | Check whether the PeopleNet engine, inference configuration, parser, and labels come from the same version. |
| USB cannot be opened | Check /dev/video*, supported formats, permissions, and device occupancy. |
| No CSI image | Check the compatible module, connection, sensor driver, Argus service, and camera occupancy. |
| MP4 missing or cannot seek | Camera mode only previews; in video mode, wait for EOS or a single Ctrl+C wrap-up before checking. |