Hugging Face's logo

qualcomm
/
DDColor

Image-to-Image
PyTorch
LiteRT
android
Model card Files
xet
Community

DDColor: Optimized for Mobile Deployment

Colorize image from the black-and-white image

DDColor is a coloring algorithm that produces natural, vivid color results from incoming black and white images.

This model is an implementation of DDColor found here.

This repository provides scripts to run DDColor on Qualcomm® devices. More details on model performance across various devices, can be found here.

Model Details

  • Model Type: Model_use_case.image_editing
  • Model Stats:
    • Model checkpoint: ddcolor_paper_tiny.pth
    • Input resolution: 224x224
    • Number of parameters: 56.3M
    • Model size (float): 215 MB
    • Model size (w8a8): 54.8 MB
Model Precision Device Chipset Target Runtime Inference Time (ms) Peak Memory Range (MB) Primary Compute Unit Target Model
DDColor float QCS8275 (Proxy) Qualcomm® QCS8275 (Proxy) TFLITE 248.698 ms 1 - 352 MB NPU DDColor.tflite
DDColor float QCS8275 (Proxy) Qualcomm® QCS8275 (Proxy) QNN_DLC 1990.164 ms 1 - 444 MB NPU DDColor.dlc
DDColor float QCS8450 (Proxy) Qualcomm® QCS8450 (Proxy) TFLITE 168.888 ms 1 - 263 MB NPU DDColor.tflite
DDColor float QCS8450 (Proxy) Qualcomm® QCS8450 (Proxy) QNN_DLC 1246.016 ms 0 - 248 MB NPU DDColor.dlc
DDColor float QCS8550 (Proxy) Qualcomm® QCS8550 (Proxy) TFLITE 159.228 ms 0 - 37 MB NPU DDColor.tflite
DDColor float QCS8550 (Proxy) Qualcomm® QCS8550 (Proxy) QNN_DLC 1122.164 ms 0 - 45 MB NPU DDColor.dlc
DDColor float QCS9075 (Proxy) Qualcomm® QCS9075 (Proxy) TFLITE 612.742 ms 1 - 354 MB NPU DDColor.tflite
DDColor float QCS9075 (Proxy) Qualcomm® QCS9075 (Proxy) QNN_DLC 1115.818 ms 0 - 710 MB NPU DDColor.dlc
DDColor float SA7255P ADP Qualcomm® SA7255P TFLITE 248.698 ms 1 - 352 MB NPU DDColor.tflite
DDColor float SA7255P ADP Qualcomm® SA7255P QNN_DLC 1990.164 ms 1 - 444 MB NPU DDColor.dlc
DDColor float SA8255 (Proxy) Qualcomm® SA8255P (Proxy) TFLITE 159.59 ms 0 - 36 MB NPU DDColor.tflite
DDColor float SA8255 (Proxy) Qualcomm® SA8255P (Proxy) QNN_DLC 1122.822 ms 0 - 43 MB NPU DDColor.dlc
DDColor float SA8295P ADP Qualcomm® SA8295P TFLITE 176.513 ms 0 - 242 MB NPU DDColor.tflite
DDColor float SA8295P ADP Qualcomm® SA8295P QNN_DLC 1232.732 ms 1 - 402 MB NPU DDColor.dlc
DDColor float SA8650 (Proxy) Qualcomm® SA8650P (Proxy) TFLITE 159.237 ms 0 - 38 MB NPU DDColor.tflite
DDColor float SA8650 (Proxy) Qualcomm® SA8650P (Proxy) QNN_DLC 1124.211 ms 0 - 41 MB NPU DDColor.dlc
DDColor float SA8775P ADP Qualcomm® SA8775P TFLITE 612.742 ms 1 - 354 MB NPU DDColor.tflite
DDColor float SA8775P ADP Qualcomm® SA8775P QNN_DLC 1115.818 ms 0 - 710 MB NPU DDColor.dlc
DDColor float Samsung Galaxy S24 Snapdragon® 8 Gen 3 Mobile TFLITE 115.014 ms 0 - 377 MB NPU DDColor.tflite
DDColor float Samsung Galaxy S24 Snapdragon® 8 Gen 3 Mobile QNN_DLC 847.726 ms 0 - 846 MB NPU DDColor.dlc
DDColor float Samsung Galaxy S25 Snapdragon® 8 Elite For Galaxy Mobile TFLITE 94.763 ms 1 - 310 MB NPU DDColor.tflite
DDColor float Samsung Galaxy S25 Snapdragon® 8 Elite For Galaxy Mobile QNN_DLC 840.905 ms 1 - 434 MB NPU DDColor.dlc
DDColor float Snapdragon 8 Elite Gen 5 QRD Snapdragon® 8 Elite Gen5 Mobile TFLITE 73.571 ms 1 - 330 MB NPU DDColor.tflite
DDColor float Snapdragon 8 Elite Gen 5 QRD Snapdragon® 8 Elite Gen5 Mobile QNN_DLC 708.725 ms 1 - 664 MB NPU DDColor.dlc
DDColor float Snapdragon X Elite CRD Snapdragon® X Elite QNN_DLC 1161.306 ms 296 - 296 MB NPU DDColor.dlc
DDColor w8a8 Dragonwing RB3 Gen 2 Vision Kit Qualcomm® QCS6490 TFLITE 653.671 ms 113 - 242 MB CPU DDColor.tflite
DDColor w8a8 QCS8275 (Proxy) Qualcomm® QCS8275 (Proxy) TFLITE 3047.28 ms 0 - 351 MB NPU DDColor.tflite
DDColor w8a8 QCS8450 (Proxy) Qualcomm® QCS8450 (Proxy) TFLITE 1700.149 ms 0 - 424 MB NPU DDColor.tflite
DDColor w8a8 QCS8550 (Proxy) Qualcomm® QCS8550 (Proxy) TFLITE 1609.53 ms 0 - 33 MB NPU DDColor.tflite
DDColor w8a8 QCS9075 (Proxy) Qualcomm® QCS9075 (Proxy) TFLITE 1631.545 ms 0 - 350 MB NPU DDColor.tflite
DDColor w8a8 RB5 (Proxy) Qualcomm® QCS8250 (Proxy) TFLITE 829.323 ms 35 - 97 MB CPU DDColor.tflite
DDColor w8a8 SA7255P ADP Qualcomm® SA7255P TFLITE 3047.28 ms 0 - 351 MB NPU DDColor.tflite
DDColor w8a8 SA8255 (Proxy) Qualcomm® SA8255P (Proxy) TFLITE 1611.308 ms 0 - 39 MB NPU DDColor.tflite
DDColor w8a8 SA8295P ADP Qualcomm® SA8295P TFLITE 1938.979 ms 0 - 399 MB NPU DDColor.tflite
DDColor w8a8 SA8650 (Proxy) Qualcomm® SA8650P (Proxy) TFLITE 1610.256 ms 0 - 22 MB NPU DDColor.tflite
DDColor w8a8 SA8775P ADP Qualcomm® SA8775P TFLITE 1631.545 ms 0 - 350 MB NPU DDColor.tflite
DDColor w8a8 Samsung Galaxy S24 Snapdragon® 8 Gen 3 Mobile TFLITE 1223.884 ms 0 - 366 MB NPU DDColor.tflite
DDColor w8a8 Samsung Galaxy S25 Snapdragon® 8 Elite For Galaxy Mobile TFLITE 1161.027 ms 0 - 350 MB NPU DDColor.tflite
DDColor w8a8 Snapdragon 7 Gen 4 QRD Snapdragon® 7 Gen 4 Mobile TFLITE 472.892 ms 105 - 355 MB CPU DDColor.tflite
DDColor w8a8 Snapdragon 8 Elite Gen 5 QRD Snapdragon® 8 Elite Gen5 Mobile TFLITE 1165.779 ms 0 - 351 MB NPU DDColor.tflite

Installation

Install the package via pip:

# NOTE: 3.10 <= PYTHON_VERSION < 3.14 is supported.
pip install "qai-hub-models[ddcolor]"

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.ddcolor.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.ddcolor.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.ddcolor.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.ddcolor 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.ddcolor.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.ddcolor.demo -- --eval-mode on-device

Deploying compiled model to Android

The models can be deployed using multiple runtimes:

  • TensorFlow Lite (.tflite export): This tutorial provides a guide to deploy the .tflite model in an Android application.

  • QNN (.so export ): This sample app provides instructions on how to use the .so shared library in an Android application.

View on Qualcomm® AI Hub

Get more details on DDColor's performance across various devices here. Explore all available models on Qualcomm® AI Hub

License

  • The license for the original implementation of DDColor can be found here.
  • The license for the compiled assets for on-device deployment can be found here

References

Community

Downloads last month
93
Inference Providers NEW
Image-to-Image
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support