--- library_name: pytorch license: other tags: - android pipeline_tag: image-segmentation --- ![](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/ffnet_78s/web-assets/model_demo.png) # FFNet-78S: Optimized for Mobile Deployment ## Semantic segmentation for automotive street scenes FFNet-78S is a "fuss-free network" that segments street scene images with per-pixel classes like road, sidewalk, and pedestrian. Trained on the Cityscapes dataset. This model is an implementation of FFNet-78S found [here](https://github.com/Qualcomm-AI-research/FFNet). This repository provides scripts to run FFNet-78S on Qualcomm® devices. More details on model performance across various devices, can be found [here](https://aihub.qualcomm.com/models/ffnet_78s). ### Model Details - **Model Type:** Model_use_case.semantic_segmentation - **Model Stats:** - Model checkpoint: ffnet78S_dBBB_cityscapes_state_dict_quarts - Input resolution: 2048x1024 - Number of output classes: 19 - Number of parameters: 27.5M - Model size (float): 105 MB - Model size (w8a8): 26.7 MB | Model | Precision | Device | Chipset | Target Runtime | Inference Time (ms) | Peak Memory Range (MB) | Primary Compute Unit | Target Model |---|---|---|---|---|---|---|---|---| | FFNet-78S | float | QCS8275 (Proxy) | Qualcomm® QCS8275 (Proxy) | TFLITE | 200.039 ms | 2 - 83 MB | NPU | [FFNet-78S.tflite](https://huggingface.co/qualcomm/FFNet-78S/blob/main/FFNet-78S.tflite) | | FFNet-78S | float | QCS8275 (Proxy) | Qualcomm® QCS8275 (Proxy) | QNN_DLC | 182.656 ms | 24 - 91 MB | NPU | [FFNet-78S.dlc](https://huggingface.co/qualcomm/FFNet-78S/blob/main/FFNet-78S.dlc) | | FFNet-78S | float | QCS8450 (Proxy) | Qualcomm® QCS8450 (Proxy) | TFLITE | 67.447 ms | 2 - 133 MB | NPU | [FFNet-78S.tflite](https://huggingface.co/qualcomm/FFNet-78S/blob/main/FFNet-78S.tflite) | | FFNet-78S | float | QCS8450 (Proxy) | Qualcomm® QCS8450 (Proxy) | QNN_DLC | 77.842 ms | 24 - 89 MB | NPU | [FFNet-78S.dlc](https://huggingface.co/qualcomm/FFNet-78S/blob/main/FFNet-78S.dlc) | | FFNet-78S | float | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | TFLITE | 51.234 ms | 2 - 20 MB | NPU | [FFNet-78S.tflite](https://huggingface.co/qualcomm/FFNet-78S/blob/main/FFNet-78S.tflite) | | FFNet-78S | float | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | QNN_DLC | 40.861 ms | 24 - 45 MB | NPU | [FFNet-78S.dlc](https://huggingface.co/qualcomm/FFNet-78S/blob/main/FFNet-78S.dlc) | | FFNet-78S | float | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | ONNX | 34.196 ms | 0 - 106 MB | NPU | [FFNet-78S.onnx.zip](https://huggingface.co/qualcomm/FFNet-78S/blob/main/FFNet-78S.onnx.zip) | | FFNet-78S | float | QCS9075 (Proxy) | Qualcomm® QCS9075 (Proxy) | TFLITE | 70.051 ms | 2 - 82 MB | NPU | [FFNet-78S.tflite](https://huggingface.co/qualcomm/FFNet-78S/blob/main/FFNet-78S.tflite) | | FFNet-78S | float | QCS9075 (Proxy) | Qualcomm® QCS9075 (Proxy) | QNN_DLC | 58.548 ms | 24 - 90 MB | NPU | [FFNet-78S.dlc](https://huggingface.co/qualcomm/FFNet-78S/blob/main/FFNet-78S.dlc) | | FFNet-78S | float | SA7255P ADP | Qualcomm® SA7255P | TFLITE | 200.039 ms | 2 - 83 MB | NPU | [FFNet-78S.tflite](https://huggingface.co/qualcomm/FFNet-78S/blob/main/FFNet-78S.tflite) | | FFNet-78S | float | SA7255P ADP | Qualcomm® SA7255P | QNN_DLC | 182.656 ms | 24 - 91 MB | NPU | [FFNet-78S.dlc](https://huggingface.co/qualcomm/FFNet-78S/blob/main/FFNet-78S.dlc) | | FFNet-78S | float | SA8255 (Proxy) | Qualcomm® SA8255P (Proxy) | TFLITE | 51.006 ms | 2 - 20 MB | NPU | [FFNet-78S.tflite](https://huggingface.co/qualcomm/FFNet-78S/blob/main/FFNet-78S.tflite) | | FFNet-78S | float | SA8255 (Proxy) | Qualcomm® SA8255P (Proxy) | QNN_DLC | 40.502 ms | 24 - 50 MB | NPU | [FFNet-78S.dlc](https://huggingface.co/qualcomm/FFNet-78S/blob/main/FFNet-78S.dlc) | | FFNet-78S | float | SA8295P ADP | Qualcomm® SA8295P | TFLITE | 76.502 ms | 2 - 81 MB | NPU | [FFNet-78S.tflite](https://huggingface.co/qualcomm/FFNet-78S/blob/main/FFNet-78S.tflite) | | FFNet-78S | float | SA8295P ADP | Qualcomm® SA8295P | QNN_DLC | 63.63 ms | 24 - 89 MB | NPU | [FFNet-78S.dlc](https://huggingface.co/qualcomm/FFNet-78S/blob/main/FFNet-78S.dlc) | | FFNet-78S | float | SA8650 (Proxy) | Qualcomm® SA8650P (Proxy) | TFLITE | 51.018 ms | 2 - 23 MB | NPU | [FFNet-78S.tflite](https://huggingface.co/qualcomm/FFNet-78S/blob/main/FFNet-78S.tflite) | | FFNet-78S | float | SA8650 (Proxy) | Qualcomm® SA8650P (Proxy) | QNN_DLC | 40.681 ms | 24 - 45 MB | NPU | [FFNet-78S.dlc](https://huggingface.co/qualcomm/FFNet-78S/blob/main/FFNet-78S.dlc) | | FFNet-78S | float | SA8775P ADP | Qualcomm® SA8775P | TFLITE | 70.051 ms | 2 - 82 MB | NPU | [FFNet-78S.tflite](https://huggingface.co/qualcomm/FFNet-78S/blob/main/FFNet-78S.tflite) | | FFNet-78S | float | SA8775P ADP | Qualcomm® SA8775P | QNN_DLC | 58.548 ms | 24 - 90 MB | NPU | [FFNet-78S.dlc](https://huggingface.co/qualcomm/FFNet-78S/blob/main/FFNet-78S.dlc) | | FFNet-78S | float | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | TFLITE | 33.979 ms | 2 - 131 MB | NPU | [FFNet-78S.tflite](https://huggingface.co/qualcomm/FFNet-78S/blob/main/FFNet-78S.tflite) | | FFNet-78S | float | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | QNN_DLC | 28.274 ms | 24 - 91 MB | NPU | [FFNet-78S.dlc](https://huggingface.co/qualcomm/FFNet-78S/blob/main/FFNet-78S.dlc) | | FFNet-78S | float | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | ONNX | 24.191 ms | 25 - 112 MB | NPU | [FFNet-78S.onnx.zip](https://huggingface.co/qualcomm/FFNet-78S/blob/main/FFNet-78S.onnx.zip) | | FFNet-78S | float | Samsung Galaxy S25 | Snapdragon® 8 Elite For Galaxy Mobile | TFLITE | 27.312 ms | 1 - 84 MB | NPU | [FFNet-78S.tflite](https://huggingface.co/qualcomm/FFNet-78S/blob/main/FFNet-78S.tflite) | | FFNet-78S | float | Samsung Galaxy S25 | Snapdragon® 8 Elite For Galaxy Mobile | QNN_DLC | 21.2 ms | 23 - 105 MB | NPU | [FFNet-78S.dlc](https://huggingface.co/qualcomm/FFNet-78S/blob/main/FFNet-78S.dlc) | | FFNet-78S | float | Samsung Galaxy S25 | Snapdragon® 8 Elite For Galaxy Mobile | ONNX | 28.974 ms | 7 - 72 MB | NPU | [FFNet-78S.onnx.zip](https://huggingface.co/qualcomm/FFNet-78S/blob/main/FFNet-78S.onnx.zip) | | FFNet-78S | float | Snapdragon 8 Elite Gen 5 QRD | Snapdragon® 8 Elite Gen5 Mobile | TFLITE | 22.457 ms | 1 - 84 MB | NPU | [FFNet-78S.tflite](https://huggingface.co/qualcomm/FFNet-78S/blob/main/FFNet-78S.tflite) | | FFNet-78S | float | Snapdragon 8 Elite Gen 5 QRD | Snapdragon® 8 Elite Gen5 Mobile | QNN_DLC | 15.777 ms | 24 - 113 MB | NPU | [FFNet-78S.dlc](https://huggingface.co/qualcomm/FFNet-78S/blob/main/FFNet-78S.dlc) | | FFNet-78S | float | Snapdragon 8 Elite Gen 5 QRD | Snapdragon® 8 Elite Gen5 Mobile | ONNX | 13.472 ms | 30 - 134 MB | NPU | [FFNet-78S.onnx.zip](https://huggingface.co/qualcomm/FFNet-78S/blob/main/FFNet-78S.onnx.zip) | | FFNet-78S | float | Snapdragon X Elite CRD | Snapdragon® X Elite | QNN_DLC | 41.752 ms | 37 - 37 MB | NPU | [FFNet-78S.dlc](https://huggingface.co/qualcomm/FFNet-78S/blob/main/FFNet-78S.dlc) | | FFNet-78S | float | Snapdragon X Elite CRD | Snapdragon® X Elite | ONNX | 33.978 ms | 31 - 31 MB | NPU | [FFNet-78S.onnx.zip](https://huggingface.co/qualcomm/FFNet-78S/blob/main/FFNet-78S.onnx.zip) | | FFNet-78S | w8a8 | Dragonwing RB3 Gen 2 Vision Kit | Qualcomm® QCS6490 | TFLITE | 51.218 ms | 1 - 36 MB | NPU | [FFNet-78S.tflite](https://huggingface.co/qualcomm/FFNet-78S/blob/main/FFNet-78S_w8a8.tflite) | | FFNet-78S | w8a8 | Dragonwing RB3 Gen 2 Vision Kit | Qualcomm® QCS6490 | QNN_DLC | 84.569 ms | 6 - 289 MB | NPU | [FFNet-78S.dlc](https://huggingface.co/qualcomm/FFNet-78S/blob/main/FFNet-78S_w8a8.dlc) | | FFNet-78S | w8a8 | Dragonwing RB3 Gen 2 Vision Kit | Qualcomm® QCS6490 | ONNX | 487.386 ms | 162 - 230 MB | CPU | [FFNet-78S.onnx.zip](https://huggingface.co/qualcomm/FFNet-78S/blob/main/FFNet-78S_w8a8.onnx.zip) | | FFNet-78S | w8a8 | QCS8275 (Proxy) | Qualcomm® QCS8275 (Proxy) | TFLITE | 29.589 ms | 1 - 56 MB | NPU | [FFNet-78S.tflite](https://huggingface.co/qualcomm/FFNet-78S/blob/main/FFNet-78S_w8a8.tflite) | | FFNet-78S | w8a8 | QCS8275 (Proxy) | Qualcomm® QCS8275 (Proxy) | QNN_DLC | 39.501 ms | 6 - 77 MB | NPU | [FFNet-78S.dlc](https://huggingface.co/qualcomm/FFNet-78S/blob/main/FFNet-78S_w8a8.dlc) | | FFNet-78S | w8a8 | QCS8450 (Proxy) | Qualcomm® QCS8450 (Proxy) | TFLITE | 12.042 ms | 1 - 84 MB | NPU | [FFNet-78S.tflite](https://huggingface.co/qualcomm/FFNet-78S/blob/main/FFNet-78S_w8a8.tflite) | | FFNet-78S | w8a8 | QCS8450 (Proxy) | Qualcomm® QCS8450 (Proxy) | QNN_DLC | 20.857 ms | 6 - 99 MB | NPU | [FFNet-78S.dlc](https://huggingface.co/qualcomm/FFNet-78S/blob/main/FFNet-78S_w8a8.dlc) | | FFNet-78S | w8a8 | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | TFLITE | 10.667 ms | 1 - 14 MB | NPU | [FFNet-78S.tflite](https://huggingface.co/qualcomm/FFNet-78S/blob/main/FFNet-78S_w8a8.tflite) | | FFNet-78S | w8a8 | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | QNN_DLC | 17.047 ms | 6 - 27 MB | NPU | [FFNet-78S.dlc](https://huggingface.co/qualcomm/FFNet-78S/blob/main/FFNet-78S_w8a8.dlc) | | FFNet-78S | w8a8 | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | ONNX | 11.655 ms | 0 - 75 MB | NPU | [FFNet-78S.onnx.zip](https://huggingface.co/qualcomm/FFNet-78S/blob/main/FFNet-78S_w8a8.onnx.zip) | | FFNet-78S | w8a8 | QCS9075 (Proxy) | Qualcomm® QCS9075 (Proxy) | TFLITE | 11.306 ms | 1 - 56 MB | NPU | [FFNet-78S.tflite](https://huggingface.co/qualcomm/FFNet-78S/blob/main/FFNet-78S_w8a8.tflite) | | FFNet-78S | w8a8 | QCS9075 (Proxy) | Qualcomm® QCS9075 (Proxy) | QNN_DLC | 17.375 ms | 6 - 77 MB | NPU | [FFNet-78S.dlc](https://huggingface.co/qualcomm/FFNet-78S/blob/main/FFNet-78S_w8a8.dlc) | | FFNet-78S | w8a8 | RB5 (Proxy) | Qualcomm® QCS8250 (Proxy) | TFLITE | 338.654 ms | 1 - 3 MB | NPU | [FFNet-78S.tflite](https://huggingface.co/qualcomm/FFNet-78S/blob/main/FFNet-78S_w8a8.tflite) | | FFNet-78S | w8a8 | RB5 (Proxy) | Qualcomm® QCS8250 (Proxy) | ONNX | 504.264 ms | 158 - 191 MB | CPU | [FFNet-78S.onnx.zip](https://huggingface.co/qualcomm/FFNet-78S/blob/main/FFNet-78S_w8a8.onnx.zip) | | FFNet-78S | w8a8 | SA7255P ADP | Qualcomm® SA7255P | TFLITE | 29.589 ms | 1 - 56 MB | NPU | [FFNet-78S.tflite](https://huggingface.co/qualcomm/FFNet-78S/blob/main/FFNet-78S_w8a8.tflite) | | FFNet-78S | w8a8 | SA7255P ADP | Qualcomm® SA7255P | QNN_DLC | 39.501 ms | 6 - 77 MB | NPU | [FFNet-78S.dlc](https://huggingface.co/qualcomm/FFNet-78S/blob/main/FFNet-78S_w8a8.dlc) | | FFNet-78S | w8a8 | SA8255 (Proxy) | Qualcomm® SA8255P (Proxy) | TFLITE | 10.695 ms | 1 - 12 MB | NPU | [FFNet-78S.tflite](https://huggingface.co/qualcomm/FFNet-78S/blob/main/FFNet-78S_w8a8.tflite) | | FFNet-78S | w8a8 | SA8255 (Proxy) | Qualcomm® SA8255P (Proxy) | QNN_DLC | 17.122 ms | 6 - 25 MB | NPU | [FFNet-78S.dlc](https://huggingface.co/qualcomm/FFNet-78S/blob/main/FFNet-78S_w8a8.dlc) | | FFNet-78S | w8a8 | SA8295P ADP | Qualcomm® SA8295P | TFLITE | 16.843 ms | 1 - 59 MB | NPU | [FFNet-78S.tflite](https://huggingface.co/qualcomm/FFNet-78S/blob/main/FFNet-78S_w8a8.tflite) | | FFNet-78S | w8a8 | SA8295P ADP | Qualcomm® SA8295P | QNN_DLC | 23.571 ms | 6 - 80 MB | NPU | [FFNet-78S.dlc](https://huggingface.co/qualcomm/FFNet-78S/blob/main/FFNet-78S_w8a8.dlc) | | FFNet-78S | w8a8 | SA8650 (Proxy) | Qualcomm® SA8650P (Proxy) | TFLITE | 10.697 ms | 1 - 12 MB | NPU | [FFNet-78S.tflite](https://huggingface.co/qualcomm/FFNet-78S/blob/main/FFNet-78S_w8a8.tflite) | | FFNet-78S | w8a8 | SA8650 (Proxy) | Qualcomm® SA8650P (Proxy) | QNN_DLC | 16.977 ms | 6 - 26 MB | NPU | [FFNet-78S.dlc](https://huggingface.co/qualcomm/FFNet-78S/blob/main/FFNet-78S_w8a8.dlc) | | FFNet-78S | w8a8 | SA8775P ADP | Qualcomm® SA8775P | TFLITE | 11.306 ms | 1 - 56 MB | NPU | [FFNet-78S.tflite](https://huggingface.co/qualcomm/FFNet-78S/blob/main/FFNet-78S_w8a8.tflite) | | FFNet-78S | w8a8 | SA8775P ADP | Qualcomm® SA8775P | QNN_DLC | 17.375 ms | 6 - 77 MB | NPU | [FFNet-78S.dlc](https://huggingface.co/qualcomm/FFNet-78S/blob/main/FFNet-78S_w8a8.dlc) | | FFNet-78S | w8a8 | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | TFLITE | 7.654 ms | 1 - 79 MB | NPU | [FFNet-78S.tflite](https://huggingface.co/qualcomm/FFNet-78S/blob/main/FFNet-78S_w8a8.tflite) | | FFNet-78S | w8a8 | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | QNN_DLC | 11.805 ms | 6 - 100 MB | NPU | [FFNet-78S.dlc](https://huggingface.co/qualcomm/FFNet-78S/blob/main/FFNet-78S_w8a8.dlc) | | FFNet-78S | w8a8 | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | ONNX | 7.982 ms | 6 - 91 MB | NPU | [FFNet-78S.onnx.zip](https://huggingface.co/qualcomm/FFNet-78S/blob/main/FFNet-78S_w8a8.onnx.zip) | | FFNet-78S | w8a8 | Samsung Galaxy S25 | Snapdragon® 8 Elite For Galaxy Mobile | TFLITE | 6.199 ms | 1 - 61 MB | NPU | [FFNet-78S.tflite](https://huggingface.co/qualcomm/FFNet-78S/blob/main/FFNet-78S_w8a8.tflite) | | FFNet-78S | w8a8 | Samsung Galaxy S25 | Snapdragon® 8 Elite For Galaxy Mobile | QNN_DLC | 7.96 ms | 6 - 87 MB | NPU | [FFNet-78S.dlc](https://huggingface.co/qualcomm/FFNet-78S/blob/main/FFNet-78S_w8a8.dlc) | | FFNet-78S | w8a8 | Samsung Galaxy S25 | Snapdragon® 8 Elite For Galaxy Mobile | ONNX | 6.173 ms | 1 - 71 MB | NPU | [FFNet-78S.onnx.zip](https://huggingface.co/qualcomm/FFNet-78S/blob/main/FFNet-78S_w8a8.onnx.zip) | | FFNet-78S | w8a8 | Snapdragon 7 Gen 4 QRD | Snapdragon® 7 Gen 4 Mobile | TFLITE | 16.656 ms | 0 - 76 MB | NPU | [FFNet-78S.tflite](https://huggingface.co/qualcomm/FFNet-78S/blob/main/FFNet-78S_w8a8.tflite) | | FFNet-78S | w8a8 | Snapdragon 7 Gen 4 QRD | Snapdragon® 7 Gen 4 Mobile | QNN_DLC | 22.082 ms | 6 - 99 MB | NPU | [FFNet-78S.dlc](https://huggingface.co/qualcomm/FFNet-78S/blob/main/FFNet-78S_w8a8.dlc) | | FFNet-78S | w8a8 | Snapdragon 7 Gen 4 QRD | Snapdragon® 7 Gen 4 Mobile | ONNX | 531.424 ms | 147 - 164 MB | CPU | [FFNet-78S.onnx.zip](https://huggingface.co/qualcomm/FFNet-78S/blob/main/FFNet-78S_w8a8.onnx.zip) | | FFNet-78S | w8a8 | Snapdragon 8 Elite Gen 5 QRD | Snapdragon® 8 Elite Gen5 Mobile | TFLITE | 5.406 ms | 1 - 61 MB | NPU | [FFNet-78S.tflite](https://huggingface.co/qualcomm/FFNet-78S/blob/main/FFNet-78S_w8a8.tflite) | | FFNet-78S | w8a8 | Snapdragon 8 Elite Gen 5 QRD | Snapdragon® 8 Elite Gen5 Mobile | QNN_DLC | 5.892 ms | 6 - 101 MB | NPU | [FFNet-78S.dlc](https://huggingface.co/qualcomm/FFNet-78S/blob/main/FFNet-78S_w8a8.dlc) | | FFNet-78S | w8a8 | Snapdragon 8 Elite Gen 5 QRD | Snapdragon® 8 Elite Gen5 Mobile | ONNX | 4.516 ms | 7 - 83 MB | NPU | [FFNet-78S.onnx.zip](https://huggingface.co/qualcomm/FFNet-78S/blob/main/FFNet-78S_w8a8.onnx.zip) | | FFNet-78S | w8a8 | Snapdragon X Elite CRD | Snapdragon® X Elite | QNN_DLC | 17.759 ms | 115 - 115 MB | NPU | [FFNet-78S.dlc](https://huggingface.co/qualcomm/FFNet-78S/blob/main/FFNet-78S_w8a8.dlc) | | FFNet-78S | w8a8 | Snapdragon X Elite CRD | Snapdragon® X Elite | ONNX | 12.165 ms | 22 - 22 MB | NPU | [FFNet-78S.onnx.zip](https://huggingface.co/qualcomm/FFNet-78S/blob/main/FFNet-78S_w8a8.onnx.zip) | ## Installation Install the package via pip: ```bash # NOTE: 3.10 <= PYTHON_VERSION < 3.14 is supported. pip install "qai-hub-models[ffnet-78s]" ``` ## Configure Qualcomm® AI Hub Workbench to run this model on a cloud-hosted device Sign-in to [Qualcomm® AI Hub Workbench](https://workbench.aihub.qualcomm.com/) with your Qualcomm® ID. Once signed in navigate to `Account -> Settings -> API Token`. With this API token, you can configure your client to run models on the cloud hosted devices. ```bash qai-hub configure --api_token API_TOKEN ``` Navigate to [docs](https://workbench.aihub.qualcomm.com/docs/) for more information. ## Demo off target The package contains a simple end-to-end demo that downloads pre-trained weights and runs this model on a sample input. ```bash python -m qai_hub_models.models.ffnet_78s.demo ``` The above demo runs a reference implementation of pre-processing, model inference, and post processing. **NOTE**: If you want running in a Jupyter Notebook or Google Colab like environment, please add the following to your cell (instead of the above). ``` %run -m qai_hub_models.models.ffnet_78s.demo ``` ### Run model on a cloud-hosted device In addition to the demo, you can also run the model on a cloud-hosted Qualcomm® device. This script does the following: * Performance check on-device on a cloud-hosted device * Downloads compiled assets that can be deployed on-device for Android. * Accuracy check between PyTorch and on-device outputs. ```bash python -m qai_hub_models.models.ffnet_78s.export ``` ## How does this work? This [export script](https://aihub.qualcomm.com/models/ffnet_78s/qai_hub_models/models/FFNet-78S/export.py) leverages [Qualcomm® AI Hub](https://aihub.qualcomm.com/) to optimize, validate, and deploy this model on-device. Lets go through each step below in detail: Step 1: **Compile model for on-device deployment** To compile a PyTorch model for on-device deployment, we first trace the model in memory using the `jit.trace` and then call the `submit_compile_job` API. ```python import torch import qai_hub as hub from qai_hub_models.models.ffnet_78s import Model # Load the model torch_model = Model.from_pretrained() # Device device = hub.Device("Samsung Galaxy S25") # Trace model input_shape = torch_model.get_input_spec() sample_inputs = torch_model.sample_inputs() pt_model = torch.jit.trace(torch_model, [torch.tensor(data[0]) for _, data in sample_inputs.items()]) # Compile model on a specific device compile_job = hub.submit_compile_job( model=pt_model, device=device, input_specs=torch_model.get_input_spec(), ) # Get target model to run on-device target_model = compile_job.get_target_model() ``` Step 2: **Performance profiling on cloud-hosted device** After compiling models from step 1. Models can be profiled model on-device using the `target_model`. Note that this scripts runs the model on a device automatically provisioned in the cloud. Once the job is submitted, you can navigate to a provided job URL to view a variety of on-device performance metrics. ```python profile_job = hub.submit_profile_job( model=target_model, device=device, ) ``` Step 3: **Verify on-device accuracy** To verify the accuracy of the model on-device, you can run on-device inference on sample input data on the same cloud hosted device. ```python input_data = torch_model.sample_inputs() inference_job = hub.submit_inference_job( model=target_model, device=device, inputs=input_data, ) on_device_output = inference_job.download_output_data() ``` With the output of the model, you can compute like PSNR, relative errors or spot check the output with expected output. **Note**: This on-device profiling and inference requires access to Qualcomm® AI Hub Workbench. [Sign up for access](https://myaccount.qualcomm.com/signup). ## Run demo on a cloud-hosted device You can also run the demo on-device. ```bash python -m qai_hub_models.models.ffnet_78s.demo --eval-mode on-device ``` **NOTE**: If you want running in a Jupyter Notebook or Google Colab like environment, please add the following to your cell (instead of the above). ``` %run -m qai_hub_models.models.ffnet_78s.demo -- --eval-mode on-device ``` ## Deploying compiled model to Android The models can be deployed using multiple runtimes: - TensorFlow Lite (`.tflite` export): [This tutorial](https://www.tensorflow.org/lite/android/quickstart) provides a guide to deploy the .tflite model in an Android application. - QNN (`.so` export ): This [sample app](https://docs.qualcomm.com/bundle/publicresource/topics/80-63442-50/sample_app.html) provides instructions on how to use the `.so` shared library in an Android application. ## View on Qualcomm® AI Hub Get more details on FFNet-78S's performance across various devices [here](https://aihub.qualcomm.com/models/ffnet_78s). Explore all available models on [Qualcomm® AI Hub](https://aihub.qualcomm.com/) ## License * The license for the original implementation of FFNet-78S can be found [here](https://github.com/Qualcomm-AI-research/FFNet/blob/master/LICENSE). * The license for the compiled assets for on-device deployment can be found [here](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/Qualcomm+AI+Hub+Proprietary+License.pdf) ## References * [Simple and Efficient Architectures for Semantic Segmentation](https://arxiv.org/abs/2206.08236) * [Source Model Implementation](https://github.com/Qualcomm-AI-research/FFNet) ## Community * Join [our AI Hub Slack community](https://aihub.qualcomm.com/community/slack) to collaborate, post questions and learn more about on-device AI. * For questions or feedback please [reach out to us](mailto:ai-hub-support@qti.qualcomm.com).