35 रिपॉजिटरी
Interoperable formats for exporting trained models to different runtimes and production environments.
Distinct from Model Inference Deployment: Focuses on the serialization and export to formats like ONNX, PMML, and CoreML, rather than the deployment orchestration.
Explore 35 awesome GitHub repositories matching devops & infrastructure · Model Export Formats. Refine with filters or upvote what's useful.
This is a real-time object detection framework built on the YOLOv3 architecture, implemented in PyTorch. It provides a complete pipeline for identifying and localizing objects in images and video using a single neural network pass, combining a Darknet-53 backbone with multi-scale feature pyramids and anchor-based bounding box prediction. The framework extends beyond basic detection to include instance segmentation, human pose estimation, and multi-object tracking across video frames. It offers a model export toolkit that converts trained models through ONNX to CoreML, TensorFlow Lite, and Ten
Exports trained models to ONNX, CoreML, TFLite, and TensorRT for production deployment.
YOLOX is a high-performance anchor-free YOLO, exceeding yolov3~v5 with MegEngine, ONNX, TensorRT, ncnn, and OpenVINO supported. Documentation: https://yolox.readthedocs.io/
Ships a dedicated export pipeline converting trained detection models to ONNX, TensorRT, ncnn, OpenVINO, and MegEngine formats.
This project is a structured learning curriculum and technical reference for mastering deep learning with TensorFlow. It provides a comprehensive guide for building, training, and deploying neural networks, combining theoretical fundamentals with practical implementation examples. The repository distinguishes itself by covering the end-to-end machine learning workflow, from low-level tensor mathematics and linear algebra to the creation of complex model architectures. It includes specific guidance on developing data pipelines for diverse data types, such as images, text, and time-series seque
Details how to export model weights and parameters into interoperable formats for deployment.
YOLOv9 is a real-time computer vision framework and deep learning model designed for image classification, object detection, and instance segmentation. It functions as both a vision model and a trainer, allowing for the optimization of neural network weights on custom datasets using single or multiple GPUs. The framework utilizes programmable gradient information to perform high-speed identification and location of multiple objects within images and video streams. It extends beyond bounding box detection to provide instance segmentation and panoptic segmentation, which labels every pixel in a
Exports trained models into multiple formats like ONNX and TensorRT for diverse hardware deployment.
CatBoost is a gradient boosting machine learning library used to train decision tree ensembles for regression, classification, and ranking tasks. It functions as a high-performance framework that provides a categorical data processor for transforming non-numeric features, a distributed trainer for large-scale datasets, and GPU acceleration to speed up model construction. The library distinguishes itself through native handling of categorical data and text features, removing the need for manual encoding. It includes a specialized model interpretability tool that leverages SHAP values and featu
Exports trained models to ONNX, PMML, or CoreML for high-speed execution in production environments or databases.
TensorFlow.js is a JavaScript machine learning library and browser-based runtime used to build, train, and execute models. It functions as a WebGL accelerated tensor engine, providing a foundation for high-performance linear algebra operations and an automatic differentiation framework for computing gradients. The project distinguishes itself through its ability to run machine learning directly in web environments, supporting both client-side inference and browser-based training. It enables the deployment of Python-based models by converting Keras or TensorFlow models into compatible formats
Imports TFLite flatbuffers from URLs or memory buffers into an interpreter for execution.
PaddleGAN is a generative AI framework and deep learning computer vision library built on the PaddlePaddle framework. It serves as a toolkit for image and video synthesis, providing a collection of generative adversarial network implementations for creating synthetic visual content. The library focuses on advanced synthesis capabilities, including the generation of talking heads through lip motion synchronization and the creation of synthetic videos via motion transfer from driving sequences. It provides tools for domain-to-domain translation, allowing for image style transfer and the transfo
Converts trained generative models into deployable formats by specifying dimensions and weights.
This project provides a suite of lightweight face detection models designed for high-speed inference on edge computing devices. It centers on a compact neural network architecture that enables human face detection within environments characterized by limited compute resources and power constraints. The system features quantized face detectors available in multiple formats to ensure compatibility across diverse hardware architectures. It includes utilities for model export and quantization, allowing trained weights to be converted into standardized formats for hardware-agnostic deployment. Th
Includes toolkits to convert trained weights into multiple standardized formats for cross-platform deployment.
h2o-3 is a distributed machine learning platform and automated machine learning framework designed for training and deploying predictive models using distributed in-memory computing. It functions as a deep learning framework and a distributed model scoring engine, capable of operating as a Kubernetes ML cluster to process large datasets in parallel. The platform distinguishes itself through automated machine learning capabilities that automatically select the best algorithms and hyperparameters to optimize model performance. It provides specialized deep learning toolkits for tasks including i
Exports trained models into serialized formats to enable high-speed predictions in production environments.
BackgroundMattingV2 is a deep learning background matting tool and real-time image segmentation framework. It provides a system for isolating foreground subjects from high-resolution images and video feeds in real time. The project includes a deep learning model trainer for optimizing matting models through base convergence and end-to-end refinement. It also functions as a cross-runtime model exporter, converting trained neural networks into interchangeable formats for deployment across different software environments and hardware runtimes. The framework supports streaming processed webcam f
Converts trained matting models into multiple interchangeable formats like ONNX for cross-platform deployment.
यह विविध नेचुरल लैंग्वेज प्रोसेसिंग कार्यों को एक एकीकृत टेक्स्ट-टू-टेक्स्ट समस्या के रूप में मानने के लिए एक मशीन लर्निंग फ्रेमवर्क है। यह बड़े पैमाने पर ट्रांसफॉर्मर मॉडल को प्री-ट्रेन और फाइन-ट्यून करने के लिए एक टूलकिट प्रदान करता है, जो एक ऐसी प्रणाली का उपयोग करता है जहां इनपुट और आउटपुट दोनों को रॉ टेक्स्ट अनुक्रमों के रूप में स्वरूपित किया जाता है। फ्रेमवर्क अपने डिस्ट्रीब्यूटेड ट्रेनिंग सिस्टम के लिए जाना जाता है, जो कई TPU कोर में मॉडल वेट्स और ट्रेनिंग बैच को स्केल करने के लिए मेश-आधारित रणनीतियों का उपयोग करता है। यह कॉन्फ़िगर करने योग्य मिश्रण दरों का उपयोग करके विविध डेटासेट को एक ही ट्रेनिंग स्ट्रीम में जोड़कर मल्टी-टास्क लर्निंग का समर्थन करता है, जिससे एक ही मॉडल विभिन्न भाषा कार्यों को संभालने में सक्षम होता है। सिस्टम एनकोडर-डिकोडर आर्किटेक्चर, टेक्स्ट जनरेशन के लिए बीम-सर्च डिकोडिंग, और ट्रांसफर लर्निंग वर्कफ़्लो सहित क्षमताओं की एक विस्तृत श्रृंखला को कवर करता है। इसमें NLP डेटासेट तैयारी, मॉडल प्रदर्शन मूल्यांकन, और प्रोडक्शन सर्विंग के लिए प्रशिक्षित चेकपॉइंट्स को एक्सपोर्ट करने के लिए उपयोगिताएँ शामिल हैं। लाइब्रेरी विकास में तेजी लाने के लिए विभिन्न आकारों के प्री-ट्रेंड मॉडल चेकपॉइंट्स को लोड करने का समर्थन करती है।
Converts trained checkpoints into deployment-ready formats suitable for production environments or container images.
ByteTrack is a multi-object tracking framework that implements the ByteTrack algorithm, an ECCV 2022 method designed to recover occluded objects and reduce trajectory fragmentation. The core innovation of the project is its association algorithm, which processes every detection box—including low-confidence ones—by using separate high and low score thresholds, Kalman filter motion prediction, and Hungarian algorithm matching to produce consistent object identities across video frames. The project distinguishes itself by its comprehensive approach to handling occlusions and fragmented trajector
Exports tracking models to ONNX, TensorRT, ncnn, and DeepStream formats for production inference.
YOLO-World is a vision-language framework and open-vocabulary object detection model. It identifies objects in images and video based on free-form text prompts without requiring predefined category labels. The system enables the identification of arbitrary objects by fusing image features with text embeddings. It includes a specialized tool for automated image labeling, which generates bounding box annotations for custom datasets using text-based prompts. The project provides a deployment pipeline for converting models into quantized ONNX and TFLite formats, supporting real-time inference on
Converts models to TFLite format using INT8 quantization for efficient mobile deployment.
This project is a collection of TensorFlow 2.x machine learning tutorials and practical code examples. It serves as a deep learning implementation guide for constructing diverse neural network architectures, including convolutional, recurrent, and generative networks. The repository provides templates and examples for several specialized domains, including computer vision for image classification and object detection, natural language processing for text generation and language understanding, and generative AI for synthesizing data using adversarial networks and autoencoders. It also includes
Provides interoperable formats for exporting trained models to different runtimes and production environments.
Darkflow is an object detection framework and computer vision pipeline that provides a programmatic interface for performing real-time image analysis and object identification. It functions as a tool for loading weights, fine-tuning models, and executing inference on both static images and video feeds. The project serves as a converter that translates Darknet configurations and weights into TensorFlow graphs to enable retraining and deployment. It includes a model exporter that saves trained graphs into portable protobuf files for use on mobile and native devices. The system covers capabilit
Saves trained graphs and weights into portable binary formats like Protobuf for deployment on mobile devices.
Chinese-CLIP एक मल्टीमॉडल फ्रेमवर्क और विज़न-लैंग्वेज मॉडल है जिसे चीनी भाषा के टेक्स्ट और इमेजेस का उपयोग करके क्रॉस-मॉडल रिट्रीवल और रिप्रेजेंटेशन जनरेशन के लिए बनाया गया है। यह विज़ुअल और टेक्स्टुअल डेटा को एक साझा वेक्टर स्पेस में मैप करने के लिए कॉन्ट्रास्टिव लर्निंग आर्किटेक्चर का उपयोग करता है। यह सिस्टम टेक्स्ट-टू-इमेज और इमेज-टू-टेक्स्ट रिट्रीवल के लिए द्वि-दिशात्मक (bidirectional) सर्च की सुविधा देता है। साथ ही, यह ज़ीरो-शॉट इमेज क्लासिफिकेशन भी प्रदान करता है, जो बिना किसी विशिष्ट ट्रेनिंग के इमेजेस में ऑब्जेक्ट्स की पहचान कर सकता है। इस प्रोजेक्ट में डिस्ट्रीब्यूटेड ट्रेनिंग और कॉन्ट्रास्टिव लर्निंग का उपयोग करके प्री-ट्रेंड मॉडल्स को फाइन-ट्यून करने के टूल्स शामिल हैं। यह इन्फरेंस स्पीड बढ़ाने के लिए मॉडल वेट्स को ऑप्टिमाइज़्ड फॉर्मेट में एक्सपोर्ट करने की सुविधा भी देता है।
Includes utilities to export model weights into optimized formats for faster production inference.
Anomalib is a PyTorch-based library for visual anomaly detection, offering a modular framework, a comprehensive model zoo, and a benchmarking suite designed for industrial defect detection. It provides a wide range of algorithms—including generative, discriminative, teacher-student, and vision-language approaches—that support unsupervised, few-shot, and zero-shot settings. The library enables deployment through model export to ONNX and OpenVINO for edge devices, and includes a no-code web application for training and inference. It also features a command-line interface for orchestrating multi
Converts trained models to ONNX, OpenVINO IR, or PyTorch for production deployment.
MMdnn is a deep learning model converter and migrator designed to translate neural network architectures and weights between different frameworks such as TensorFlow, PyTorch, and Keras. It utilizes a standardized intermediate representation to decouple network structures and weights from specific framework implementations, enabling the transformation of pre-trained models across different environments. The project distinguishes itself by generating native Python reconstruction code from its intermediate representations, allowing models to be rebuilt and fine-tuned in target environments. It a
Synthesizes original framework-specific model binaries by combining generated architecture code and binary weight data.
RF-DETR is a Python library for training and deploying object detection, instance segmentation, and keypoint detection models built on a vision transformer architecture. It provides a unified command-line interface and Python API for the full workflow, from fine-tuning pretrained checkpoints on custom datasets to running inference on images, video files, and live camera streams. The project supports training on datasets in COCO or YOLO format, with automatic format detection and configurable augmentation pipelines. Models can be exported to ONNX, TFLite, or TensorRT for deployment across edge
Converts trained detection models to ONNX and TFLite for production deployment.
Composer एक PyTorch डिस्ट्रिब्यूटेड ट्रेनिंग फ्रेमवर्क है जिसे मल्टी-नोड GPU क्लस्टर में बड़े पैमाने के मॉडल को स्केल करने के लिए डिज़ाइन किया गया है। यह एक लार्ज लैंग्वेज मॉडल ट्रेनर, डिस्ट्रिब्यूटेड मॉडल ऑप्टिमाइज़र और ट्रेनिंग लाइफसाइकिल मैनेजर के रूप में कार्य करता है। यह प्रोजेक्ट एक डीप लर्निंग रेगुलराइजेशन लाइब्रेरी के रूप में खुद को अलग करता है, जो मॉडल सामान्यीकरण (generalization) को बेहतर बनाने के लिए Sharpness Aware Minimization, MixUp और CutMix जैसी विशेष ऑप्टिमाइज़ेशन तकनीकें प्रदान करता है। यह सीक्वेंस लेंथ वार्मअप, प्रोग्रेसिव लेयर फ्रीजिंग और बड़े पैमाने पर मॉडल रिकवरी के लिए शार्डेड-स्टेट चेकपॉइंटिंग के उपयोग के माध्यम से अपने ट्रेनिंग फ्लो को और अलग करता है। यह फ्रेमवर्क डिस्ट्रिब्यूटेड ट्रेनिंग ऑर्केस्ट्रेशन, मिक्स्ड-प्रिसिजन हार्डवेयर मैनेजमेंट और क्लाउड-नेटिव डेटा स्ट्रीमिंग सहित क्षमताओं की एक विस्तृत श्रृंखला को कवर करता है। यह GPU मेमोरी डायग्नोस्टिक्स, ट्रेनिंग डाइवर्जेंस डिटेक्शन और थ्रूपुट ट्रैकिंग के लिए व्यापक मॉनिटरिंग और ऑब्जर्वेबिलिटी टूल भी प्रदान करता है। इस प्रोजेक्ट में नोड्स के बीच मल्टी-GPU ट्रेनिंग जॉब्स के निष्पादन को स्वचालित करने के लिए एक कमांड-लाइन लॉन्चर शामिल है।
Converts trained models into portable formats optimized for production deployment and storage.