A collection of Jupyter notebooks demonstrating TensorFlow Lite model quantization, conversion, and optimization techniques for deep neural networks.
Adventures in TensorFlow Lite is a collection of Jupyter notebooks that demonstrate how to optimize and deploy deep learning models using TensorFlow Lite. It provides practical examples of quantization techniques, model conversion workflows, and optimization strategies to make neural networks efficient for edge and mobile devices. The repository covers various computer vision models including segmentation, style transfer, and text detection.
Machine learning engineers and developers who need to deploy TensorFlow models to resource-constrained environments like mobile devices or edge hardware. It's particularly useful for those working on model optimization, quantization, and TensorFlow Lite conversion.
This repository offers hands-on, practical examples with real-world models rather than theoretical explanations. It covers both quantization-aware training and post-training quantization, includes Edge TPU compatibility, and demonstrates complete workflows from model conversion to inference.
This repository contains notebooks that show the usage of TensorFlow Lite for quantizing deep neural networks.
Open-Awesome is built by the community, for the community. Submit a project, suggest an awesome list, or help improve the catalog on GitHub.
Provides Jupyter notebooks with complete code for quantizing and converting real-world models like DeepLabV3 and ESRGAN, making optimization techniques accessible through immediate experimentation.
Includes both quantization-aware training and post-training quantization with fine-tuning, as demonstrated in the 'A_tale_of_quantization.ipynb' notebook, backed by a detailed report.
Shows how to prepare models for Google Edge TPU acceleration, with examples in 'Custom_Image_Classification_EdgeTPU.ipynb' for calibrating models with representative datasets.
Covers multiple models and tasks, such as segmentation, style transfer, and text detection, offering a broad range of optimization scenarios for edge deployment.
Admits to being under active development with potential inconsistencies in the README, which can lead to bugs or outdated information for users relying on stable examples.
The PyTorch to TensorFlow Lite conversion workflow in 'TUNIT_Conversion_to_TF_Lite.ipynb' is explicitly noted as buggy, reducing its reliability for such tasks.
Primarily focuses on computer vision models, with limited examples for other domains like audio or NLP, narrowing its applicability beyond visual tasks.