Code of the ICASSP 2022 paper "Gradient Variance Loss for Structure Enhanced Super-Resolution"

本仓库包含了完整的深度学习应用开发流程，以经典的手写字符识别为例，基于LeNet网络构建。推理部分使用torch、onnxruntime以及openvino框架💖

Vision-lanugage model example code.

DA2Lite is an automated model compression toolkit for PyTorch.

Minimal Reproducibility Study of (https://arxiv.org/abs/1911.05248). Experiments with Compression of Deep Neural Networks

ptdeco is a library for model optimization by matrix decomposition built on top of PyTorch

quantization example for pqt & qat

compares different pretrained object classification with per-layer and per-channel quantization using pytorch

Quantization for Object Detection in Tensorflow 2.x

Computer vision project that classifies 101 food categories with 80.2% accuracy using fine-tuned EfficientNetB2 and PyTorch. Features interactive Gradio UI, optimized inference (~100ms/image), and strategic training on 20% of Food101 dataset for efficient resource utilization.

Công cụ giảm kích thước mô hình bằng Quantization, kết hợp AI Agent để tự động chọn mức tối ưu, giúp tăng tốc và tiết kiệm chi phí inference.

ai-zipper offers numerous AI model compression methods, also it is easy to embed into your own source code

This project builds and optimizes a model on a dataset using Ridge regression and polynomial features. Model accuracy is enhanced through regularization and polynomial transformations. Grid search and cross-validation are used to find the best parameters, and the model's performance is evaluated.

MedMNIST-EdgeAI -> an end-to-end exploration into model distillation, optimization, and deployment for resource-constrained environments, all centered around the MedMNIST medical imaging dataset.

Heart disease classification using machine learning algorithms with hyperparameter tuning for optimized model performance. Algorithms include XGBoost, Random Forest, Logistic Regression, and moreto find the best model for accurate heart disease prediction.

The objective of this project is the development and evaluation of recommendation algorithms based on the MovieLens dataset, one of the benchmark datasets for research into recommendation systems. User ratings, tags, and movie metadata are used in the dataset, allowing for simple and advanced recommendation techniques

This project focuses on real-time object detection and tracking using the Faster R-CNN model, emphasizing accuracy over speed. It utilizes the COCO 2017 dataset for training, which contains diverse and complex images. The Faster R-CNN model is integrated with FiftyOne for visualizing predictions and ground truth annotations. A custom CentroidTracke

People Counter App at the Edge