Deep Residual Learning for Image Recognition
Paper
•
1512.03385
•
Published
•
12
ResNet101: Optimized for Mobile Deployment
Imagenet classifier and general purpose backbone
ResNet101 is a machine learning model that can classify images from the Imagenet dataset. It can also be used as a backbone in building more complex models for specific use cases.
This model is an implementation of ResNet101 found here.
This repository provides scripts to run ResNet101 on Qualcomm® devices. More details on model performance across various devices, can be found here.
Model Details
- Model Type: Model_use_case.image_classification
- Model Stats:
- Model checkpoint: Imagenet
- Input resolution: 224x224
- Number of parameters: 44.5M
- Model size (float): 170 MB
- Model size (w8a8): 43.9 MB
| Model | Precision | Device | Chipset | Target Runtime | Inference Time (ms) | Peak Memory Range (MB) | Primary Compute Unit | Target Model |
|---|---|---|---|---|---|---|---|---|
| ResNet101 | float | QCS8275 (Proxy) | Qualcomm® QCS8275 (Proxy) | TFLITE | 18.336 ms | 0 - 202 MB | NPU | ResNet101.tflite |
| ResNet101 | float | QCS8275 (Proxy) | Qualcomm® QCS8275 (Proxy) | QNN_DLC | 18.266 ms | 1 - 150 MB | NPU | ResNet101.dlc |
| ResNet101 | float | QCS8450 (Proxy) | Qualcomm® QCS8450 (Proxy) | TFLITE | 5.808 ms | 0 - 247 MB | NPU | ResNet101.tflite |
| ResNet101 | float | QCS8450 (Proxy) | Qualcomm® QCS8450 (Proxy) | QNN_DLC | 5.833 ms | 1 - 186 MB | NPU | ResNet101.dlc |
| ResNet101 | float | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | TFLITE | 3.277 ms | 0 - 3 MB | NPU | ResNet101.tflite |
| ResNet101 | float | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | QNN_DLC | 3.277 ms | 1 - 3 MB | NPU | ResNet101.dlc |
| ResNet101 | float | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | ONNX | 3.337 ms | 0 - 98 MB | NPU | ResNet101.onnx.zip |
| ResNet101 | float | QCS9075 (Proxy) | Qualcomm® QCS9075 (Proxy) | TFLITE | 5.424 ms | 0 - 201 MB | NPU | ResNet101.tflite |
| ResNet101 | float | QCS9075 (Proxy) | Qualcomm® QCS9075 (Proxy) | QNN_DLC | 23.494 ms | 1 - 150 MB | NPU | ResNet101.dlc |
| ResNet101 | float | SA7255P ADP | Qualcomm® SA7255P | TFLITE | 18.336 ms | 0 - 202 MB | NPU | ResNet101.tflite |
| ResNet101 | float | SA7255P ADP | Qualcomm® SA7255P | QNN_DLC | 18.266 ms | 1 - 150 MB | NPU | ResNet101.dlc |
| ResNet101 | float | SA8255 (Proxy) | Qualcomm® SA8255P (Proxy) | TFLITE | 3.264 ms | 0 - 3 MB | NPU | ResNet101.tflite |
| ResNet101 | float | SA8255 (Proxy) | Qualcomm® SA8255P (Proxy) | QNN_DLC | 3.266 ms | 1 - 3 MB | NPU | ResNet101.dlc |
| ResNet101 | float | SA8295P ADP | Qualcomm® SA8295P | TFLITE | 5.503 ms | 0 - 185 MB | NPU | ResNet101.tflite |
| ResNet101 | float | SA8295P ADP | Qualcomm® SA8295P | QNN_DLC | 5.53 ms | 1 - 135 MB | NPU | ResNet101.dlc |
| ResNet101 | float | SA8650 (Proxy) | Qualcomm® SA8650P (Proxy) | TFLITE | 3.245 ms | 0 - 3 MB | NPU | ResNet101.tflite |
| ResNet101 | float | SA8650 (Proxy) | Qualcomm® SA8650P (Proxy) | QNN_DLC | 3.285 ms | 1 - 3 MB | NPU | ResNet101.dlc |
| ResNet101 | float | SA8775P ADP | Qualcomm® SA8775P | TFLITE | 5.424 ms | 0 - 201 MB | NPU | ResNet101.tflite |
| ResNet101 | float | SA8775P ADP | Qualcomm® SA8775P | QNN_DLC | 23.494 ms | 1 - 150 MB | NPU | ResNet101.dlc |
| ResNet101 | float | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | TFLITE | 2.351 ms | 0 - 264 MB | NPU | ResNet101.tflite |
| ResNet101 | float | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | QNN_DLC | 2.377 ms | 1 - 208 MB | NPU | ResNet101.dlc |
| ResNet101 | float | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | ONNX | 2.391 ms | 0 - 166 MB | NPU | ResNet101.onnx.zip |
| ResNet101 | float | Samsung Galaxy S25 | Snapdragon® 8 Elite For Galaxy Mobile | TFLITE | 1.937 ms | 0 - 204 MB | NPU | ResNet101.tflite |
| ResNet101 | float | Samsung Galaxy S25 | Snapdragon® 8 Elite For Galaxy Mobile | QNN_DLC | 1.94 ms | 1 - 154 MB | NPU | ResNet101.dlc |
| ResNet101 | float | Samsung Galaxy S25 | Snapdragon® 8 Elite For Galaxy Mobile | ONNX | 2.022 ms | 0 - 123 MB | NPU | ResNet101.onnx.zip |
| ResNet101 | float | Snapdragon 8 Elite Gen 5 QRD | Snapdragon® 8 Elite Gen5 Mobile | TFLITE | 1.636 ms | 0 - 204 MB | NPU | ResNet101.tflite |
| ResNet101 | float | Snapdragon 8 Elite Gen 5 QRD | Snapdragon® 8 Elite Gen5 Mobile | QNN_DLC | 1.653 ms | 1 - 155 MB | NPU | ResNet101.dlc |
| ResNet101 | float | Snapdragon 8 Elite Gen 5 QRD | Snapdragon® 8 Elite Gen5 Mobile | ONNX | 1.768 ms | 0 - 125 MB | NPU | ResNet101.onnx.zip |
| ResNet101 | float | Snapdragon X Elite CRD | Snapdragon® X Elite | QNN_DLC | 3.506 ms | 1 - 1 MB | NPU | ResNet101.dlc |
| ResNet101 | float | Snapdragon X Elite CRD | Snapdragon® X Elite | ONNX | 3.247 ms | 86 - 86 MB | NPU | ResNet101.onnx.zip |
| ResNet101 | w8a8 | Dragonwing Q-6690 MTP | Qualcomm® Qcm6690 | TFLITE | 11.87 ms | 0 - 175 MB | NPU | ResNet101.tflite |
| ResNet101 | w8a8 | Dragonwing Q-6690 MTP | Qualcomm® Qcm6690 | QNN_DLC | 13.632 ms | 0 - 177 MB | NPU | ResNet101.dlc |
| ResNet101 | w8a8 | Dragonwing Q-6690 MTP | Qualcomm® Qcm6690 | ONNX | 41.665 ms | 8 - 24 MB | CPU | ResNet101.onnx.zip |
| ResNet101 | w8a8 | Dragonwing RB3 Gen 2 Vision Kit | Qualcomm® QCS6490 | TFLITE | 4.306 ms | 0 - 45 MB | NPU | ResNet101.tflite |
| ResNet101 | w8a8 | Dragonwing RB3 Gen 2 Vision Kit | Qualcomm® QCS6490 | QNN_DLC | 4.565 ms | 0 - 2 MB | NPU | ResNet101.dlc |
| ResNet101 | w8a8 | Dragonwing RB3 Gen 2 Vision Kit | Qualcomm® QCS6490 | ONNX | 56.379 ms | 8 - 58 MB | CPU | ResNet101.onnx.zip |
| ResNet101 | w8a8 | QCS8275 (Proxy) | Qualcomm® QCS8275 (Proxy) | TFLITE | 2.762 ms | 0 - 162 MB | NPU | ResNet101.tflite |
| ResNet101 | w8a8 | QCS8275 (Proxy) | Qualcomm® QCS8275 (Proxy) | QNN_DLC | 2.953 ms | 0 - 163 MB | NPU | ResNet101.dlc |
| ResNet101 | w8a8 | QCS8450 (Proxy) | Qualcomm® QCS8450 (Proxy) | TFLITE | 1.65 ms | 0 - 228 MB | NPU | ResNet101.tflite |
| ResNet101 | w8a8 | QCS8450 (Proxy) | Qualcomm® QCS8450 (Proxy) | QNN_DLC | 1.798 ms | 0 - 216 MB | NPU | ResNet101.dlc |
| ResNet101 | w8a8 | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | TFLITE | 1.129 ms | 0 - 3 MB | NPU | ResNet101.tflite |
| ResNet101 | w8a8 | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | QNN_DLC | 1.282 ms | 0 - 2 MB | NPU | ResNet101.dlc |
| ResNet101 | w8a8 | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | ONNX | 1.5 ms | 0 - 51 MB | NPU | ResNet101.onnx.zip |
| ResNet101 | w8a8 | QCS9075 (Proxy) | Qualcomm® QCS9075 (Proxy) | TFLITE | 1.333 ms | 0 - 161 MB | NPU | ResNet101.tflite |
| ResNet101 | w8a8 | QCS9075 (Proxy) | Qualcomm® QCS9075 (Proxy) | QNN_DLC | 5.233 ms | 0 - 163 MB | NPU | ResNet101.dlc |
| ResNet101 | w8a8 | RB5 (Proxy) | Qualcomm® QCS8250 (Proxy) | TFLITE | 45.431 ms | 0 - 186 MB | GPU | ResNet101.tflite |
| ResNet101 | w8a8 | RB5 (Proxy) | Qualcomm® QCS8250 (Proxy) | ONNX | 38.797 ms | 9 - 47 MB | CPU | ResNet101.onnx.zip |
| ResNet101 | w8a8 | SA7255P ADP | Qualcomm® SA7255P | TFLITE | 2.762 ms | 0 - 162 MB | NPU | ResNet101.tflite |
| ResNet101 | w8a8 | SA7255P ADP | Qualcomm® SA7255P | QNN_DLC | 2.953 ms | 0 - 163 MB | NPU | ResNet101.dlc |
| ResNet101 | w8a8 | SA8255 (Proxy) | Qualcomm® SA8255P (Proxy) | TFLITE | 1.133 ms | 0 - 3 MB | NPU | ResNet101.tflite |
| ResNet101 | w8a8 | SA8255 (Proxy) | Qualcomm® SA8255P (Proxy) | QNN_DLC | 1.271 ms | 0 - 3 MB | NPU | ResNet101.dlc |
| ResNet101 | w8a8 | SA8295P ADP | Qualcomm® SA8295P | TFLITE | 1.76 ms | 0 - 169 MB | NPU | ResNet101.tflite |
| ResNet101 | w8a8 | SA8295P ADP | Qualcomm® SA8295P | QNN_DLC | 1.909 ms | 0 - 169 MB | NPU | ResNet101.dlc |
| ResNet101 | w8a8 | SA8650 (Proxy) | Qualcomm® SA8650P (Proxy) | TFLITE | 1.127 ms | 0 - 2 MB | NPU | ResNet101.tflite |
| ResNet101 | w8a8 | SA8650 (Proxy) | Qualcomm® SA8650P (Proxy) | QNN_DLC | 1.313 ms | 0 - 2 MB | NPU | ResNet101.dlc |
| ResNet101 | w8a8 | SA8775P ADP | Qualcomm® SA8775P | TFLITE | 1.333 ms | 0 - 161 MB | NPU | ResNet101.tflite |
| ResNet101 | w8a8 | SA8775P ADP | Qualcomm® SA8775P | QNN_DLC | 5.233 ms | 0 - 163 MB | NPU | ResNet101.dlc |
| ResNet101 | w8a8 | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | TFLITE | 0.868 ms | 0 - 226 MB | NPU | ResNet101.tflite |
| ResNet101 | w8a8 | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | QNN_DLC | 0.969 ms | 0 - 219 MB | NPU | ResNet101.dlc |
| ResNet101 | w8a8 | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | ONNX | 1.11 ms | 0 - 209 MB | NPU | ResNet101.onnx.zip |
| ResNet101 | w8a8 | Samsung Galaxy S25 | Snapdragon® 8 Elite For Galaxy Mobile | TFLITE | 0.742 ms | 0 - 163 MB | NPU | ResNet101.tflite |
| ResNet101 | w8a8 | Samsung Galaxy S25 | Snapdragon® 8 Elite For Galaxy Mobile | QNN_DLC | 0.784 ms | 0 - 166 MB | NPU | ResNet101.dlc |
| ResNet101 | w8a8 | Samsung Galaxy S25 | Snapdragon® 8 Elite For Galaxy Mobile | ONNX | 0.948 ms | 0 - 146 MB | NPU | ResNet101.onnx.zip |
| ResNet101 | w8a8 | Snapdragon 7 Gen 4 QRD | Snapdragon® 7 Gen 4 Mobile | TFLITE | 1.67 ms | 0 - 173 MB | NPU | ResNet101.tflite |
| ResNet101 | w8a8 | Snapdragon 7 Gen 4 QRD | Snapdragon® 7 Gen 4 Mobile | QNN_DLC | 1.806 ms | 0 - 178 MB | NPU | ResNet101.dlc |
| ResNet101 | w8a8 | Snapdragon 7 Gen 4 QRD | Snapdragon® 7 Gen 4 Mobile | ONNX | 39.586 ms | 8 - 25 MB | CPU | ResNet101.onnx.zip |
| ResNet101 | w8a8 | Snapdragon 8 Elite Gen 5 QRD | Snapdragon® 8 Elite Gen5 Mobile | TFLITE | 0.702 ms | 0 - 165 MB | NPU | ResNet101.tflite |
| ResNet101 | w8a8 | Snapdragon 8 Elite Gen 5 QRD | Snapdragon® 8 Elite Gen5 Mobile | QNN_DLC | 0.743 ms | 0 - 166 MB | NPU | ResNet101.dlc |
| ResNet101 | w8a8 | Snapdragon 8 Elite Gen 5 QRD | Snapdragon® 8 Elite Gen5 Mobile | ONNX | 0.926 ms | 0 - 144 MB | NPU | ResNet101.onnx.zip |
| ResNet101 | w8a8 | Snapdragon X Elite CRD | Snapdragon® X Elite | QNN_DLC | 1.276 ms | 0 - 0 MB | NPU | ResNet101.dlc |
| ResNet101 | w8a8 | Snapdragon X Elite CRD | Snapdragon® X Elite | ONNX | 1.268 ms | 44 - 44 MB | NPU | ResNet101.onnx.zip |
Installation
Install the package via pip:
pip install qai-hub-models
Configure Qualcomm® AI Hub Workbench to run this model on a cloud-hosted device
Sign-in to Qualcomm® AI Hub Workbench 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.
qai-hub configure --api_token API_TOKEN
Navigate to 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.
python -m qai_hub_models.models.resnet101.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.resnet101.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.
python -m qai_hub_models.models.resnet101.export
How does this work?
This export script leverages Qualcomm® AI Hub 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.
import torch
import qai_hub as hub
from qai_hub_models.models.resnet101 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.
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.
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.
Run demo on a cloud-hosted device
You can also run the demo on-device.
python -m qai_hub_models.models.resnet101.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.resnet101.demo -- --eval-mode on-device
Deploying compiled model to Android
The models can be deployed using multiple runtimes:
TensorFlow Lite (
.tfliteexport): This tutorial provides a guide to deploy the .tflite model in an Android application.QNN (
.soexport ): This sample app provides instructions on how to use the.soshared library in an Android application.
View on Qualcomm® AI Hub
Get more details on ResNet101's performance across various devices here. Explore all available models on Qualcomm® AI Hub
License
- The license for the original implementation of ResNet101 can be found here.
References
Community
- Join our AI Hub Slack community to collaborate, post questions and learn more about on-device AI.
- For questions or feedback please reach out to us.
- Downloads last month
- 116