DeepLabV3
DeepLabV3 is a classic semantic segmentation model proposed by Google. By extensively exploring atrous convolution, it enlarges the receptive field without increasing the parameter count, effectively addressing the loss of spatial information when segmenting multi-scale objects in deep neural networks.
- Key features: Excellent at capturing complex and fine-grained edges in images, with strong multi-scale perception. It enables pixel-level accurate class assignment and is widely used in medical imaging analysis, autonomous-driving perception, and satellite image processing.
- Version notes: This example uses the DeepLabV3 architecture. As a benchmark model in semantic segmentation, it improves recognition accuracy for objects of different sizes using an enhanced ASPP module and global average pooling. While maintaining deep feature extraction, it reconstructs spatial structures through a well-designed architecture, making it a mature and well-balanced choice for industrial use.
Environment setup
You need to set up the environment in advance.
Quick start
Download model files
O6 / O6N
cd ai_model_hub_25_Q3/models/ComputeVision/Semantic_Segmentation/onnx_deeplab_v3
wget https://www.modelscope.cn/models/cix/ai_model_hub_25_Q3/resolve/master/models/ComputeVision/Semantic_Segmentation/onnx_deeplab_v3/deeplab_v3.cix
Test the model
info
Activate the virtual environment before running.
O6 / O6N
python3 inference_npu.py
Full conversion workflow
Download model files
Linux PC
cd ai_model_hub_25_Q3/models/ComputeVision/Semantic_Segmentation/onnx_deeplab_v3/model
wget https://www.modelscope.cn/models/cix/ai_model_hub_25_Q3/resolve/master/models/ComputeVision/Semantic_Segmentation/onnx_deeplab_v3/model/deeplabv3_resnet50.onnx
Project structure
├── cfg
├── datasets
├── deeplab_v3.cix
├── inference_npu.py
├── inference_onnx.py
├── model
├── ReadMe.md
├── test_data
└── Tutorials.ipynb
Quantize and convert the model
Linux PC
cd ..
cixbuild cfg/onnx_deeplab_v3_build.cfg
Copy to device
After conversion, copy the .cix model files to the device.
Test inference on the host
Run the inference script
Linux PC
python3 inference_onnx.py --image_path test_data --onnx_path model/deeplabv3_resnet50.onnx
Inference output
Linux PC
$ python3 inference_onnx.py --image_path test_data --onnx_path model/deeplabv3_resnet50.onnx
save output: onnx_ILSVRC2012_val_00004704.JPEG
Deploy on NPU
Run the inference script
O6 / O6N
python3 inference_npu.py --image_path test_data --model_path deeplab_v3.cix
Inference output
O6 / O6N
$ python3 inference_npu.py --image_path test_data --model_path deeplab_v3.cix
npu: noe_init_context success
npu: noe_load_graph success
Input tensor count is 1.
Output tensor count is 1.
npu: noe_create_job success
save output: npu_ILSVRC2012_val_00004704.JPEG
npu: noe_clean_job success
npu: noe_unload_graph success
npu: noe_deinit_context success
