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. This model is capable of enhancing low-light images upto a great extent.
Model training code and pre-trained weights are provided by Soumik through this repository.
TensorFlow Lite model (dynamic-range quantized)
Original model
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 to TensorFlow Lite models. Metadata makes it easier for mobile developers to integrate the TensorFlow Lite models in their applications.
- Dynamic shape (contains dynamic-range and fp16 quantized models)
- Fixed shape metadata-populated models on TensorFlow Hub
Pixel 4 was used in order to run the benchmarking tests. Also, fixed-shape TensorFlow Lite models (accepting 400x400x3 images) were only benchmarked.
If you would run inference with an NVIDIA GPU-enabled environment then please follow along with this notebook - 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 of running the TensorRT optimized model on a low-light video.