# AI/ML sample applications

Source: [https://docs.qualcomm.com/doc/80-70017-50/topic/ai-ml-sample-applications.html](https://docs.qualcomm.com/doc/80-70017-50/topic/ai-ml-sample-applications.html)

Run custom use cases that show how to use the Qualcomm Linux AI/ML features using
        Qualcomm Neural Processing SDK, Qualcomm AI Engine direct and LiteRT (formerly TensorFlow
        Lite or TFLite) models.

Before you run the AI/ML sample applications, use the following procedures to download
            the model and label files on the device.

## Download model, label, and config JSON files for Qualcomm Neural Processing SDK 

To download and push the model files for QCS6490, QCS9075, and QCS8275, do the
                following on the Linux host:
1. Enable SSH and connect to Wi-Fi. For instructions, see [Log in using SSH](https://docs.qualcomm.com/bundle/publicresource/topics/80-70017-254/how_to.html#use-ssh).
Note: If SSH is already enabled and Wi-Fi is connected,
                            you can skip this step.
2. Log into the Linux host device using SSH and change directory to
                            `/opt`:

        ssh root@<ip address of device>Copy to clipboard

        cd /optCopy to clipboard
3. On the target device, obtain the
                            `download_artifacts.sh` script, set executable
                        permissions, and run it to download the model, media, and label
                        files:

        cd /tmp/Copy to clipboard

        curl -L -O https://raw.githubusercontent.com/quic/sample-apps-for-qualcomm-linux/refs/heads/main/qualcomm-linux/scripts/download_artifacts.shCopy to clipboard

        chmod +x download_artifacts.shCopy to clipboard

        ./download_artifacts.shCopy to clipboard

## Download model, label and config JSON files for LiteRT and Qualcomm AI Engine
                direct

1. Download the models from [IoT– Qualcomm AI Hub](https://aihub.qualcomm.com/iot/models/). 
The YOLOv8
                        and YOLO-NAS models are not available by default. Export these models using
                        the AI Hub APIs. For instructions, see the following:
    - [YOLOv8-Detection-Quantized](https://github.com/quic/ai-hub-models/tree/main/qai_hub_models/models/yolov8_det_quantized)
    - [Yolo-NAS-Quantized](https://github.com/quic/ai-hub-models/tree/main/qai_hub_models/models/yolonas_quantized)

    You can generate a batch model. To change the batch size of the
                        model, update <var class="keyword varname">&lt;N&gt;</var> in the following
                            `export`
                    command:

        python -m qai_hub_models.models.<Model_Name>.export --batch-size <N> --device "QCS6490 (Proxy)"Copy to clipboard
2. Update the q\_offset and q\_scale
                    constants of the quantized LiteRT model in the JSON file. For instructions, see
                        [Obtain model constants](https://docs.qualcomm.com/bundle/publicresource/topics/80-70017-15B/integrate-ai-hub-models.html#obtain-model-constants).
3. Push the model files to the device:
    1. Enable SSH to log into the host device. For instructions, see [Log in using SSH](https://docs.qualcomm.com/bundle/publicresource/topics/80-70017-254/how_to.html#use-ssh).
        If you
                                have already enabled SSH, you can skip this step.
    2. Use the following command to push the downloaded model files on the
                                device:

            scp <model filename> root@<IP addr of the target device>:/opt/Copy to clipboard

        For
                                example:

            scp mobilenet_v2_quantized.tflite root@<IP addr of the target device>:/opt/Copy to clipboard

Note: If you want to run sample applications from
                                the UART shell, remount the file system with read/write permission
                                using the following
                                command:

            mount -o remount,rw /usrCopy to clipboard
4. On the target device, obtain the `download_artifacts.sh` script,
                    set executable permissions, and run it to download the model, media, and label
                    files:

        cd /tmp/Copy to clipboard

        curl -L -O https://raw.githubusercontent.com/quic/sample-apps-for-qualcomm-linux/refs/heads/main/qualcomm-linux/scripts/download_artifacts.shCopy to clipboard

        chmod +x download_artifacts.shCopy to clipboard

        ./download_artifacts.shCopy to clipboard
5. Note: These commands download both the model and label
                        files. The model files are for Qualcomm Neural Processing SDK only. For
                        LiteRT and Qualcomm AI Engine direct, use the label files to run the
                        corresponding applications.

The following sample applications can be built using the
            default models provided by Qualcomm. If you want to *Bring Your Own Model*, see
                [AI Developer Workflow](https://docs.qualcomm.com/bundle/publicresource/topics/80-70017-15B/ai-ml-developer-workflow.html).

- **[Classification](https://docs.qualcomm.com/doc/80-70017-50/topic/gst-ai-classification.html)**  

The **gst-ai-classification** application enables you to identify the subject in         an image. The use cases use the Qualcomm Neural Processing SDK, LiteRT, or Qualcomm AI         Engine direct models.
- **[Object detection](https://docs.qualcomm.com/doc/80-70017-50/topic/gst-ai-object-detection.html)**  

The **gst-ai-object-detection** application enables you to detect objects within         images and videos. The use cases show the execution of [YOLOv5](https://github.com/ultralytics/yolov5), [YOLOv8](https://github.com/ultralytics/ultralytics), and [YOLO-NAS](https://github.com/Deci-AI/super-gradients/blob/master/YOLONAS.md) using the Qualcomm Neural Processing SDK runtime,         YOLOv8 using Qualcomm AI Engine direct, and YOLOv5 and YOLOv8 using LiteRT.
- **[Pose detection](https://docs.qualcomm.com/doc/80-70017-50/topic/gst-ai-pose-detection.html)**  

 The **gst-ai-pose-detection** application enables you to detect the body pose of         the subject in an image or video. The use cases use an input stream from a camera, file, or         an RTSP source, use LiteRT and Qualcomm AI Engine direct models for pose detection, and         display the results on the screen.
- **[Image segmentation](https://docs.qualcomm.com/doc/80-70017-50/topic/gst-ai-segmentation.html)**  

The **gst-ai-segmentation** application enables you to divide an image into         different and meaningful parts or segments and assign a label to each homogenous segment         based on the similarity of the attributes. The application shows how to use Qualcomm Neural         Processing SDK runtime, Qualcomm AI Engine direct runtime, and LiteRT for image         segmentation.
- **[Parallel AI fusion](https://docs.qualcomm.com/doc/80-70017-50/topic/gst-ai-parallel-inference.html)**  

The **gst-ai-parallel-inference** application enables you to perform object         detection, object classification, pose detection, and image segmentation on an input stream         from different sources such as a camera, a file, or an RTSP network. The use cases use the         LiteRT models for object detection, image segmentation, classification and pose         detection.
- **[Multi-input AI inferencing](https://docs.qualcomm.com/doc/80-70017-50/topic/gst-ai-multi-input-output-object-detection.html)**  

The **gst-ai-multi-input-output-object-detection** application enables you to         perform object detection on multiple video streams from different sources such as a camera,         a file, or over a network such as RTSP.
- **[Daisy chain detection and classification](https://docs.qualcomm.com/doc/80-70017-50/topic/daisy-chain-detection-and-classification.html)**  

The **gst-ai-daisychain-detection-classification** application enables you to         perform cascaded object detection and classification with a camera,  file source, or RTSP         stream. The use case involves detecting objects and classifying the detected         objects.
- **[Daisy chain detection and pose estimation](https://docs.qualcomm.com/doc/80-70017-50/topic/daisy-chain-detection-and-pose-detection.html)**  

The **gst-ai-daisychain-detection-pose** application enables you to perform         cascaded object detection and pose detection with a camera, file source, or an RTSP stream.         The use cases involve detecting objects and estimating the body poses of the subject in an         image or a video.
- **[Mono depth from video](https://docs.qualcomm.com/doc/80-70017-50/topic/mono-depth-from-video.html)**  

The **gst-ai-monodepth** application enables you to infer depth of a source feed         from a live camera stream, file, or an RTSP stream.
- **[Video super resolution](https://docs.qualcomm.com/doc/80-70017-50/topic/video-super-resolution.html)**  

The **gst-ai-superresolution** application enables you to generate high resolution         video frames from low-resolution input.
- **[Multistream inference](https://docs.qualcomm.com/doc/80-70017-50/topic/multistream-inference.html)**  

The **gst-ai-multistream-inference** application shows AI inference (object         detection and classification) on up to 16 input streams coming from camera, file, or RTSP         stream.
- **[Multistream batch inference](https://docs.qualcomm.com/doc/80-70017-50/topic/multistream-batch-inference.html)**  

The **gst-ai-multistream-batch-inference** application shows batched AI inference         (object detection and segmentation) on up to 24 input streams from video files.
- **[AI smart codec](https://docs.qualcomm.com/doc/80-70017-50/topic/ai-smart-codec.html)**  

The **gst-ai-smartcodec-example** application reduces network bandwidth and         storage resources for input from a camera or a file source.
- **[Face detection](https://docs.qualcomm.com/doc/80-70017-50/topic/gst-ai-face-detection.html)**  

The **gst-ai-face-detection** application collects the live video input from a         camera, file, or an RTSP stream and uses the Qualcomm AI Engine direct face detection model         to produce a preview with the overlaid AI model output on the HDMI display.
- **[Face recognition](https://docs.qualcomm.com/doc/80-70017-50/topic/gst-ai-face-recognition.html)**  

The **gst-ai-face-recognition** application collects the live video input from a         camera or an RTSP stream and shares this input for face detection and recognition. It uses         the three\_dmm Qualcomm AI Engine direct models for face detection. The         result is a preview of the overlaid AI model on an HDMI display.
- **[Image segmentation using Python with container](https://docs.qualcomm.com/doc/80-70017-50/topic/image-segmentation-using-python.html)**  

The application enables you to perform image segmentation using the Qualcomm Neural         Processing SDK with Python bindings, all from within a docker container.

**Parent Topic:** [Sample applications](https://docs.qualcomm.com/doc/80-70017-50/topic/example-applications.html)

Last Published: Oct 01, 2026

[Previous Topic
Sample applications](https://docs.qualcomm.com/bundle/publicresource/80-70017-50/topics/example-applications.md) [Next Topic
Classification](https://docs.qualcomm.com/bundle/publicresource/80-70017-50/topics/gst-ai-classification.md)