https://github.com/sayakpaul/mirnet-tflite-trt

TensorFlow Lite models for MIRNet for low-light image enhancement.
https://github.com/sayakpaul/mirnet-tflite-trt

computer-vision tensorflow2 tflite tinyml

Last synced: 13 days ago
JSON representation

TensorFlow Lite models for MIRNet for low-light image enhancement.

• Host: GitHub
• URL: https://github.com/sayakpaul/mirnet-tflite-trt
• Owner: sayakpaul
• Created: 2020-12-05T04:50:42.000Z (almost 4 years ago)
• Default Branch: main
• Last Pushed: 2020-12-24T02:16:20.000Z (almost 4 years ago)
• Last Synced: 2024-10-04T10:57:45.956Z (about 1 month ago)
• Topics: computer-vision, tensorflow2, tflite, tinyml
• Language: Jupyter Notebook
• Homepage:
• Size: 25.3 MB
• Stars: 129
• Watchers: 6
• Forks: 25
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

        
# MIRNet-TFLite

This repository shows the TensorFlow Lite and TensorRT model conversion and inference processes for the **MIRNet** model as proposed by [Learning Enriched Features for Real Image Restoration and Enhancement](https://arxiv.org/pdf/2003.06792v2.pdf). This model is capable of enhancing low-light images upto a great extent. 






Source



Model training code and pre-trained weights are provided by **Soumik** through [this repository](https://github.com/soumik12345/MIRNet/). 

## Comparison between the TensorFlow Lite and original models

**TensorFlow Lite model (dynamic-range quantized)**







  

**Original model**








## About the notebooks

* `MIRNet_TFLite.ipynb`: Shows the model conversion and inference processes. Models converted in this notebook support dynamic shaped inputs.  

* `MIRNet_TFLite_Fixed_Shape.ipynb`: Shows the model conversion and inference processes. Models converted in this notebook only support fixed shaped inputs. 

* `MIRNet_TRT.ipynb`: Shows the model conversion process with TensorRT as well as the inference. Recommended if you would run inference with an NVIDIA GPU-enabled environment.

* `Add_Metadata.ipynb`: Adds [metadata](https://www.tensorflow.org/lite/convert/metadata) to TensorFlow Lite models. Metadata makes it easier for mobile developers to integrate the TensorFlow Lite models in their applications. 

## TensorFlow Lite models

* [Dynamic shape](https://github.com/sayakpaul/MIRNet-TFLite/releases/download/v0.1.0/dynamic_shape.zip) (contains dynamic-range and fp16 quantized models)

* [Fixed shape metadata-populated models on TensorFlow Hub](https://tfhub.dev/sayakpaul/lite-model/mirnet-fixed/dr/1)

## Benchmarking



  



Pixel 4 was used in order to run the benchmarking tests. Also, fixed-shape TensorFlow Lite models (accepting 400x400x3 images) were only benchmarked.  



  



## Notes

If you would run inference with an NVIDIA GPU-enabled environment then please follow along with this notebook - [`MIRNet_TRT.ipynb`](https://github.com/sayakpaul/MIRNet-TFLite/blob/main/MIRNet_TRT.ipynb). If you use the TensorRT optimized model (as shown in that notebook) with an NVIDIA GPU-enabled environment the inference latency greatly improves (~0.6 seconds on a Tesla T4). [Here's a demo](https://youtu.be/BdJT3u71EDo) of running the TensorRT optimized model on a low-light video.