Semantic Segmentation

TFLite Runtime

No. Model Name Description

1 DeepLabV3-Plus-MobileNet Atrous separable convolutions with encoder-decoder for segmentation.

Steps to Run on QIMSDK

🟢 W8A8 — Model Precision

Download Required Files

File	Save as
Model	deeplabv3_plus_mobilenet_w8a8.tflite
dv3-argmax.json	dv3-argmax.json
Input video	ai_demo_sample.mp4

If any downloaded file is a .zip archive, extract it on your host machine before copying: unzip filename.zip

Copy Files to Device

# Replace $HOME to the appropriate device path before running the commands.
# For QLI:    /root
# For Ubuntu: /home/ubuntu
# Modify this based on your platform and ensure files are copied to the correct location on the device.

ssh <user>@<device-ip> "mkdir -p $HOME/{models,labels,media,media/output}"
scp deeplabv3_plus_mobilenet_w8a8.tflite   <user>@<device-ip>:$HOME/models/
scp dv3-argmax.json                        <user>@<device-ip>:$HOME/labels/
scp ai_demo_sample.mp4                     <user>@<device-ip>:$HOME/media/

Connect to device

# Run from your host machine — replace <user> and <device-ip>
ssh <user>@<device-ip>

Set Environment Variables

Qualcomm Linux

export MODEL_NAME=deeplabv3_plus_mobilenet_w8a8.tflite
export LABELS_NAME=dv3-argmax.json
export SRC_VIDEO_NAME=ai_demo_sample.mp4

Run the Pipeline

gst-launch-1.0 -e --gst-debug=2 \
filesrc location=$HOME/media/$SRC_VIDEO_NAME ! qtdemux ! h264parse ! \
v4l2h264dec capture-io-mode=4 output-io-mode=4 ! video/x-raw,format=NV12 ! queue ! \
tee name=t \
t. ! queue ! qtivcomposer name=seg_mix sink_1::alpha=0.5 ! queue ! waylandsink fullscreen=true sync=true \
t. ! queue ! qtimlvconverter ! queue ! \
qtimltflite model=$HOME/models/$MODEL_NAME delegate=external external-delegate-path=libQnnTFLiteDelegate.so external-delegate-options="QNNExternalDelegate,backend_type=htp,log_level=(string)1;" ! queue ! \
qtimlpostprocess module=deeplab-argmax labels=$HOME/labels/$LABELS_NAME ! video/x-raw,format=BGRA,width=520,height=520 ! queue ! seg_mix.

2 FCN-ResNet50 Fully convolutional network for dense pixel classification.

Steps to Run on QIMSDK

🟢 W8A8 — Model Precision

Download Required Files

File	Save as
Model	fcn_resnet50_w8a8.tflite
dv3-argmax.json	dv3-argmax.json
Input video	ai_demo_sample.mp4

If any downloaded file is a .zip archive, extract it on your host machine before copying: unzip filename.zip

Copy Files to Device

# Replace $HOME to the appropriate device path before running the commands.
# For QLI:    /root
# For Ubuntu: /home/ubuntu
# Modify this based on your platform and ensure files are copied to the correct location on the device.

ssh <user>@<device-ip> "mkdir -p $HOME/{models,labels,media,media/output}"
scp fcn_resnet50_w8a8.tflite                <user>@<device-ip>:$HOME/models/
scp dv3-argmax.json                         <user>@<device-ip>:$HOME/labels/
scp ai_demo_sample.mp4                      <user>@<device-ip>:$HOME/media/

Connect to device

# Run from your host machine — replace <user> and <device-ip>
ssh <user>@<device-ip>

Set Environment Variables

Qualcomm Linux

export MODEL_NAME=fcn_resnet50_w8a8.tflite
export LABELS_NAME=dv3-argmax.json
export SRC_VIDEO_NAME=ai_demo_sample.mp4

Run the Pipeline

gst-launch-1.0 -e --gst-debug=2 \
filesrc location=$HOME/media/$SRC_VIDEO_NAME ! qtdemux ! h264parse ! \
v4l2h264dec capture-io-mode=4 output-io-mode=4 ! video/x-raw,format=NV12 ! queue ! \
tee name=t \
t. ! queue ! qtivcomposer name=seg_mix sink_1::alpha=0.5 ! queue ! waylandsink fullscreen=true sync=true \
t. ! queue ! qtimlvconverter ! queue ! \
qtimltflite model=$HOME/models/$MODEL_NAME delegate=external external-delegate-path=libQnnTFLiteDelegate.so external-delegate-options="QNNExternalDelegate,backend_type=htp,log_level=(string)1;" ! queue ! \
qtimlpostprocess module=deeplab-argmax labels=$HOME/labels/$LABELS_NAME ! video/x-raw,format=BGRA,width=520,height=520 ! queue ! seg_mix.

3 FFNet-122NS-LowRes Fast Fusion Network for real-time semantic segmentation.

Steps to Run on QIMSDK

🟢 W8A8 — Model Precision

Download Required Files

File	Save as
Model	ffnet_122ns_lowres_w8a8.tflite
dv3-argmax.json	dv3-argmax.json
Input video	ai_demo_sample.mp4

If any downloaded file is a .zip archive, extract it on your host machine before copying: unzip filename.zip

Copy Files to Device

# Replace $HOME to the appropriate device path before running the commands.
# For QLI:    /root
# For Ubuntu: /home/ubuntu
# Modify this based on your platform and ensure files are copied to the correct location on the device.

ssh <user>@<device-ip> "mkdir -p $HOME/{models,labels,media,media/output}"
scp ffnet_122ns_lowres_w8a8.tflite      <user>@<device-ip>:$HOME/models/
scp dv3-argmax.json                     <user>@<device-ip>:$HOME/labels/
scp ai_demo_sample.mp4                  <user>@<device-ip>:$HOME/media/

Connect to device

# Run from your host machine — replace <user> and <device-ip>
ssh <user>@<device-ip>

Set Environment Variables

Qualcomm Linux

export MODEL_NAME=ffnet_122ns_lowres_w8a8.tflite
export LABELS_NAME=dv3-argmax.json
export SRC_VIDEO_NAME=ai_demo_sample.mp4

Run the Pipeline

gst-launch-1.0 -e --gst-debug=2 \
filesrc location=$HOME/media/$SRC_VIDEO_NAME ! qtdemux ! h264parse ! \
v4l2h264dec capture-io-mode=4 output-io-mode=4 ! video/x-raw,format=NV12 ! queue ! \
tee name=t \
t. ! queue ! qtivcomposer name=seg_mix sink_1::alpha=0.5 ! queue ! waylandsink fullscreen=true sync=true \
t. ! queue ! qtimlvconverter ! queue ! \
qtimltflite model=$HOME/models/$MODEL_NAME delegate=external external-delegate-path=libQnnTFLiteDelegate.so external-delegate-options="QNNExternalDelegate,backend_type=htp,log_level=(string)1;" ! queue ! \
qtimlpostprocess module=deeplab-argmax labels=$HOME/labels/$LABELS_NAME ! video/x-raw,format=BGRA,width=520,height=520 ! queue ! seg_mix.

4 FFNet-78S-LowRes Compact FFNet variant balancing speed and accuracy.

Steps to Run on QIMSDK

🟢 W8A8 — Model Precision

Download Required Files

File	Save as
Model	ffnet_78s_lowres_w8a8.tflite
dv3-argmax.json	dv3-argmax.json
Input video	ai_demo_sample.mp4

If any downloaded file is a .zip archive, extract it on your host machine before copying: unzip filename.zip

Copy Files to Device

# Replace $HOME to the appropriate device path before running the commands.
# For QLI:    /root
# For Ubuntu: /home/ubuntu
# Modify this based on your platform and ensure files are copied to the correct location on the device.

ssh <user>@<device-ip> "mkdir -p $HOME/{models,labels,media,media/output}"
scp ffnet_78s_lowres_w8a8.tflite        <user>@<device-ip>:$HOME/models/
scp dv3-argmax.json                     <user>@<device-ip>:$HOME/labels/
scp ai_demo_sample.mp4                  <user>@<device-ip>:$HOME/media/

Connect to device

# Run from your host machine — replace <user> and <device-ip>
ssh <user>@<device-ip>

Set Environment Variables

Qualcomm Linux

export MODEL_NAME=ffnet_78s_lowres_w8a8.tflite
export LABELS_NAME=dv3-argmax.json
export SRC_VIDEO_NAME=ai_demo_sample.mp4

Run the Pipeline

gst-launch-1.0 -e --gst-debug=2 \
filesrc location=$HOME/media/$SRC_VIDEO_NAME ! qtdemux ! h264parse ! \
v4l2h264dec capture-io-mode=4 output-io-mode=4 ! video/x-raw,format=NV12 ! queue ! \
tee name=t \
t. ! queue ! qtivcomposer name=seg_mix sink_1::alpha=0.5 ! queue ! waylandsink fullscreen=true sync=true \
t. ! queue ! qtimlvconverter ! queue ! \
qtimltflite model=$HOME/models/$MODEL_NAME delegate=external external-delegate-path=libQnnTFLiteDelegate.so external-delegate-options="QNNExternalDelegate,backend_type=htp,log_level=(string)1;" ! queue ! \
qtimlpostprocess module=deeplab-argmax labels=$HOME/labels/$LABELS_NAME ! video/x-raw,format=BGRA,width=520,height=520 ! queue ! seg_mix.

5 PidNet Parallel information distillation network for real-time segmentation.

Steps to Run on QIMSDK

🟢 W8A8 — Model Precision

Download Required Files

File	Save as
Model	pidnet_w8a8.tflite
dv3-argmax.json	dv3-argmax.json
Input video	ai_demo_sample.mp4

If any downloaded file is a .zip archive, extract it on your host machine before copying: unzip filename.zip

Copy Files to Device

# Replace $HOME to the appropriate device path before running the commands.
# For QLI:    /root
# For Ubuntu: /home/ubuntu
# Modify this based on your platform and ensure files are copied to the correct location on the device.

ssh <user>@<device-ip> "mkdir -p $HOME/{models,labels,media,media/output}"
scp pidnet_w8a8.tflite                  <user>@<device-ip>:$HOME/models/
scp dv3-argmax.json                     <user>@<device-ip>:$HOME/labels/
scp ai_demo_sample.mp4                  <user>@<device-ip>:$HOME/media/

Connect to device

# Run from your host machine — replace <user> and <device-ip>
ssh <user>@<device-ip>

Set Environment Variables

Qualcomm Linux

export MODEL_NAME=pidnet_w8a8.tflite
export LABELS_NAME=dv3-argmax.json
export SRC_VIDEO_NAME=ai_demo_sample.mp4

Run the Pipeline

gst-launch-1.0 -e --gst-debug=2 \
filesrc location=$HOME/media/$SRC_VIDEO_NAME ! qtdemux ! h264parse ! \
v4l2h264dec capture-io-mode=4 output-io-mode=4 ! video/x-raw,format=NV12 ! queue ! \
tee name=t \
t. ! queue ! qtivcomposer name=seg_mix sink_1::alpha=0.5 ! queue ! waylandsink fullscreen=true sync=true \
t. ! queue ! qtimlvconverter ! queue ! \
qtimltflite model=$HOME/models/$MODEL_NAME delegate=external external-delegate-path=libQnnTFLiteDelegate.so external-delegate-options="QNNExternalDelegate,backend_type=htp,log_level=(string)1;" ! queue ! \
qtimlpostprocess module=deeplab-argmax labels=$HOME/labels/$LABELS_NAME ! video/x-raw,format=BGRA,width=520,height=520 ! queue ! seg_mix.

6 DDRNet23-Slim Dual-resolution discrepancy network (slim) for fast segmentation.

Steps to Run on QIMSDK

🟢 Float32 — Model Precision

Download Required Files

File	Save as
Model	ddrnet23_slim_float.tflite
dv3-argmax.json	dv3-argmax.json
Input video	ai_demo_sample.mp4

If any downloaded file is a .zip archive, extract it on your host machine before copying: unzip filename.zip

Copy Files to Device

# Replace $HOME to the appropriate device path before running the commands.
# For QLI:    /root
# For Ubuntu: /home/ubuntu
# Modify this based on your platform and ensure files are copied to the correct location on the device.

ssh <user>@<device-ip> "mkdir -p $HOME/{models,labels,media,media/output}"
scp ddrnet23_slim_float.tflite          <user>@<device-ip>:$HOME/models/
scp dv3-argmax.json                     <user>@<device-ip>:$HOME/labels/
scp ai_demo_sample.mp4                  <user>@<device-ip>:$HOME/media/

Connect to device

# Run from your host machine — replace <user> and <device-ip>
ssh <user>@<device-ip>

Set Environment Variables

Qualcomm Linux

export MODEL_NAME=ddrnet23_slim_float.tflite
export LABELS_NAME=dv3-argmax.json
export SRC_VIDEO_NAME=ai_demo_sample.mp4

Run the Pipeline

gst-launch-1.0 -e --gst-debug=2 \
filesrc location=$HOME/media/$SRC_VIDEO_NAME ! qtdemux ! h264parse ! \
v4l2h264dec capture-io-mode=4 output-io-mode=4 ! video/x-raw,format=NV12 ! queue ! \
tee name=t \
t. ! queue ! qtivcomposer name=seg_mix sink_1::alpha=0.5 ! queue ! waylandsink fullscreen=true sync=true \
t. ! queue ! qtimlvconverter ! queue ! \
qtimltflite model=$HOME/models/$MODEL_NAME delegate=gpu ! queue ! \
qtimlpostprocess module=deeplab-argmax labels=$HOME/labels/$LABELS_NAME ! video/x-raw,format=BGRA,width=520,height=520 ! queue ! seg_mix.

7 FFNet-40S Minimal FFNet for maximum throughput.

Steps to Run on QIMSDK

🟢 W8A8 — Model Precision

Download Required Files

File	Save as
Model	ffnet_40s_w8a8.tflite
dv3-argmax.json	dv3-argmax.json
Input video	ai_demo_sample.mp4

If any downloaded file is a .zip archive, extract it on your host machine before copying: unzip filename.zip

Copy Files to Device

# Replace $HOME to the appropriate device path before running the commands.
# For QLI:    /root
# For Ubuntu: /home/ubuntu
# Modify this based on your platform and ensure files are copied to the correct location on the device.

ssh <user>@<device-ip> "mkdir -p $HOME/{models,labels,media,media/output}"
scp ffnet_40s_w8a8.tflite          <user>@<device-ip>:$HOME/models/
scp dv3-argmax.json                <user>@<device-ip>:$HOME/labels/
scp ai_demo_sample.mp4             <user>@<device-ip>:$HOME/media/

Connect to device

# Run from your host machine — replace <user> and <device-ip>
ssh <user>@<device-ip>

Set Environment Variables

Qualcomm Linux

export MODEL_NAME=ffnet_40s_w8a8.tflite
export LABELS_NAME=dv3-argmax.json
export SRC_VIDEO_NAME=ai_demo_sample.mp4

Run the Pipeline

gst-launch-1.0 -e --gst-debug=2 \
filesrc location=$HOME/media/$SRC_VIDEO_NAME ! qtdemux ! h264parse ! \
v4l2h264dec capture-io-mode=4 output-io-mode=4 ! video/x-raw,format=NV12 ! queue ! \
tee name=t \
t. ! queue ! qtivcomposer name=seg_mix sink_1::alpha=0.5 ! queue ! waylandsink fullscreen=true sync=true \
t. ! queue ! qtimlvconverter ! queue ! \
qtimltflite model=$HOME/models/$MODEL_NAME delegate=external external-delegate-path=libQnnTFLiteDelegate.so external-delegate-options="QNNExternalDelegate,backend_type=htp,log_level=(string)1;" ! queue ! \
qtimlpostprocess module=deeplab-argmax labels=$HOME/labels/$LABELS_NAME ! video/x-raw,format=BGRA,width=520,height=520 ! queue ! seg_mix.

8 FFNet-54S Small FFNet for low-latency segmentation.

Steps to Run on QIMSDK

🟢 W8A8 — Model Precision

Download Required Files

File	Save as
Model	ffnet_54s_w8a8.tflite
dv3-argmax.json	dv3-argmax.json
Input video	ai_demo_sample.mp4

If any downloaded file is a .zip archive, extract it on your host machine before copying: unzip filename.zip

Copy Files to Device

# Replace $HOME to the appropriate device path before running the commands.
# For QLI:    /root
# For Ubuntu: /home/ubuntu
# Modify this based on your platform and ensure files are copied to the correct location on the device.

ssh <user>@<device-ip> "mkdir -p $HOME/{models,labels,media,media/output}"
scp ffnet_54s_w8a8.tflite          <user>@<device-ip>:$HOME/models/
scp dv3-argmax.json                <user>@<device-ip>:$HOME/labels/
scp ai_demo_sample.mp4             <user>@<device-ip>:$HOME/media/

Connect to device

# Run from your host machine — replace <user> and <device-ip>
ssh <user>@<device-ip>

Set Environment Variables

Qualcomm Linux

export MODEL_NAME=ffnet_54s_w8a8.tflite
export LABELS_NAME=dv3-argmax.json
export SRC_VIDEO_NAME=ai_demo_sample.mp4

Run the Pipeline

gst-launch-1.0 -e --gst-debug=2 \
filesrc location=$HOME/media/$SRC_VIDEO_NAME ! qtdemux ! h264parse ! \
v4l2h264dec capture-io-mode=4 output-io-mode=4 ! video/x-raw,format=NV12 ! queue ! \
tee name=t \
t. ! queue ! qtivcomposer name=seg_mix sink_1::alpha=0.5 ! queue ! waylandsink fullscreen=true sync=true \
t. ! queue ! qtimlvconverter ! queue ! \
qtimltflite model=$HOME/models/$MODEL_NAME delegate=external external-delegate-path=libQnnTFLiteDelegate.so external-delegate-options="QNNExternalDelegate,backend_type=htp,log_level=(string)1;" ! queue ! \
qtimlpostprocess module=deeplab-argmax labels=$HOME/labels/$LABELS_NAME ! video/x-raw,format=BGRA,width=520,height=520 ! queue ! seg_mix.

9 FFNet-78S Standard-resolution FFNet-78S segmentation model.

Steps to Run on QIMSDK

🟢 W8A8 — Model Precision

Download Required Files

File	Save as
Model	ffnet_78s_w8a8.tflite
dv3-argmax.json	dv3-argmax.json
Input video	ai_demo_sample.mp4

If any downloaded file is a .zip archive, extract it on your host machine before copying: unzip filename.zip

Copy Files to Device

# Replace $HOME to the appropriate device path before running the commands.
# For QLI:    /root
# For Ubuntu: /home/ubuntu
# Modify this based on your platform and ensure files are copied to the correct location on the device.

ssh <user>@<device-ip> "mkdir -p $HOME/{models,labels,media,media/output}"
scp ffnet_78s_w8a8.tflite          <user>@<device-ip>:$HOME/models/
scp dv3-argmax.json                <user>@<device-ip>:$HOME/labels/
scp ai_demo_sample.mp4             <user>@<device-ip>:$HOME/media/

Connect to device

# Run from your host machine — replace <user> and <device-ip>
ssh <user>@<device-ip>

Set Environment Variables

Qualcomm Linux

export MODEL_NAME=ffnet_78s_w8a8.tflite
export LABELS_NAME=dv3-argmax.json
export SRC_VIDEO_NAME=ai_demo_sample.mp4

Run the Pipeline

gst-launch-1.0 -e --gst-debug=2 \
filesrc location=$HOME/media/$SRC_VIDEO_NAME ! qtdemux ! h264parse ! \
v4l2h264dec capture-io-mode=4 output-io-mode=4 ! video/x-raw,format=NV12 ! queue ! \
tee name=t \
t. ! queue ! qtivcomposer name=seg_mix sink_1::alpha=0.5 ! queue ! waylandsink fullscreen=true sync=true \
t. ! queue ! qtimlvconverter ! queue ! \
qtimltflite model=$HOME/models/$MODEL_NAME delegate=external external-delegate-path=libQnnTFLiteDelegate.so external-delegate-options="QNNExternalDelegate,backend_type=htp,log_level=(string)1;" ! queue ! \
qtimlpostprocess module=deeplab-argmax labels=$HOME/labels/$LABELS_NAME ! video/x-raw,format=BGRA,width=520,height=520 ! queue ! seg_mix.

10 Segformer-Base Hierarchical transformer with lightweight MLP decoder.

Steps to Run on QIMSDK

🟢 W8A8 — Model Precision

Download Required Files

File	Save as
Model	segformer_base_w8a8.tflite
segformer.json	segformer.json
Driving_1080p_180s_30FPS.MOV	Driving_1080p_180s_30FPS.MOV

If any downloaded file is a .zip archive, extract it on your host machine before copying: unzip filename.zip

Copy Files to Device

# Replace $HOME to the appropriate device path before running the commands.
# For QLI:    /root
# For Ubuntu: /home/ubuntu
# Modify this based on your platform and ensure files are copied to the correct location on the device.

ssh <user>@<device-ip> "mkdir -p $HOME/{models,labels,media,media/output}"
scp segformer_base_w8a8.tflite         <user>@<device-ip>:$HOME/models/
scp segformer.json                     <user>@<device-ip>:$HOME/labels/
scp Driving_1080p_180s_30FPS.MOV       <user>@<device-ip>:$HOME/media/

Connect to device

# Run from your host machine — replace <user> and <device-ip>
ssh <user>@<device-ip>

Set Environment Variables

Qualcomm Linux

export MODEL_NAME=segformer_base_w8a8.tflite
export LABELS_NAME=segformer.json
export SRC_VIDEO_NAME=Driving_1080p_180s_30FPS.MOV

Run the Pipeline

gst-launch-1.0 -e --gst-debug=2 \
filesrc location=$HOME/media/$SRC_VIDEO_NAME ! qtdemux ! h264parse ! \
v4l2h264dec capture-io-mode=4 output-io-mode=4 ! video/x-raw,format=NV12 ! queue ! \
tee name=t \
t. ! queue ! qtivcomposer name=seg_mix sink_1::alpha=0.5 ! queue ! waylandsink fullscreen=true sync=true \
t. ! queue ! qtimlvconverter ! queue ! \
qtimltflite model=$HOME/models/$MODEL_NAME delegate=external external-delegate-path=libQnnTFLiteDelegate.so external-delegate-options="QNNExternalDelegate,backend_type=htp,log_level=(string)1;" ! queue ! \
qtimlpostprocess module=deeplab-argmax labels=$HOME/labels/$LABELS_NAME ! video/x-raw,format=BGRA,width=520,height=520 ! queue ! seg_mix.

11 SINet Salient instance network for scene understanding.

Steps to Run on QIMSDK

🟢 Float32 — Model Precision

Download Required Files

File	Save as
Model	sinet_float.tflite
dv3-argmax.json	dv3-argmax.json
Input video	ai_demo_sample.mp4

If any downloaded file is a .zip archive, extract it on your host machine before copying: unzip filename.zip

Copy Files to Device

# Replace $HOME to the appropriate device path before running the commands.
# For QLI:    /root
# For Ubuntu: /home/ubuntu
# Modify this based on your platform and ensure files are copied to the correct location on the device.

ssh <user>@<device-ip> "mkdir -p $HOME/{models,labels,media,media/output}"
scp sinet_float.tflite             <user>@<device-ip>:$HOME/models/
scp dv3-argmax.json                <user>@<device-ip>:$HOME/labels/
scp ai_demo_sample.mp4             <user>@<device-ip>:$HOME/media/

Connect to device

# Run from your host machine — replace <user> and <device-ip>
ssh <user>@<device-ip>

Set Environment Variables

Qualcomm Linux

export MODEL_NAME=sinet_float.tflite
export LABELS_NAME=dv3-argmax.json
export SRC_VIDEO_NAME=ai_demo_sample.mp4

Run the Pipeline

gst-launch-1.0 -e --gst-debug=2 \
filesrc location=$HOME/media/$SRC_VIDEO_NAME ! qtdemux ! h264parse ! \
v4l2h264dec capture-io-mode=4 output-io-mode=4 ! video/x-raw,format=NV12 ! queue ! \
tee name=t \
t. ! queue ! qtivcomposer name=seg_mix sink_1::alpha=0.5 ! queue ! waylandsink fullscreen=true sync=true \
t. ! queue ! qtimlvconverter ! queue ! \
qtimltflite model=$HOME/models/$MODEL_NAME delegate=gpu ! queue ! \
qtimlpostprocess module=deeplab-argmax labels=$HOME/labels/$LABELS_NAME ! video/x-raw,format=BGRA,width=520,height=520 ! queue ! seg_mix.

12 PSPNet Pyramid pooling module for global scene context aggregation.

Steps to Run on QIMSDK

🟢 W8A8 — Model Precision

Download Required Files

File	Save as
Model	pspnet_w8a8.tflite
dv3-argmax.json	dv3-argmax.json
Input video	ai_demo_sample.mp4

If any downloaded file is a .zip archive, extract it on your host machine before copying: unzip filename.zip

Copy Files to Device

# Replace $HOME to the appropriate device path before running the commands.
# For QLI:    /root
# For Ubuntu: /home/ubuntu
# Modify this based on your platform and ensure files are copied to the correct location on the device.

ssh <user>@<device-ip> "mkdir -p $HOME/{models,labels,media,media/output}"
scp pspnet_w8a8.tflite                  <user>@<device-ip>:$HOME/models/
scp dv3-argmax.json                     <user>@<device-ip>:$HOME/labels/
scp ai_demo_sample.mp4                  <user>@<device-ip>:$HOME/media/

Connect to device

# Run from your host machine — replace <user> and <device-ip>
ssh <user>@<device-ip>

Set Environment Variables

Qualcomm Linux

export MODEL_NAME=pspnet_w8a8.tflite
export LABELS_NAME=dv3-argmax.json
export SRC_VIDEO_NAME=ai_demo_sample.mp4

Run the Pipeline

gst-launch-1.0 -e --gst-debug=2 \
filesrc location=$HOME/media/$SRC_VIDEO_NAME ! qtdemux ! h264parse ! \
v4l2h264dec capture-io-mode=4 output-io-mode=4 ! video/x-raw,format=NV12 ! queue ! \
tee name=t \
t. ! queue ! qtivcomposer name=seg_mix sink_1::alpha=0.5 ! queue ! waylandsink fullscreen=true sync=true \
t. ! queue ! qtimlvconverter ! queue ! \
qtimltflite model=$HOME/models/$MODEL_NAME delegate=external external-delegate-path=libQnnTFLiteDelegate.so external-delegate-options="QNNExternalDelegate,backend_type=htp,log_level=(string)1;" ! queue ! \
qtimlpostprocess module=deeplab-argmax labels=$HOME/labels/$LABELS_NAME ! video/x-raw,format=BGRA,width=520,height=520 ! queue ! seg_mix.

Blog

Overview

Getting Started

Building AI Pipelines

Examples

Reference

Advanced Topics