# Mobile optimization

An ultra-lightweight English text line recognition model developed by the PaddleOCR team, supporting the recognition of English and numeric characters

An ultra-lightweight Korean text line recognition model that supports the recognition of Korean and numeric characters, with an average accuracy of 60.21%.

An ultra-lightweight Cyrillic text line recognition model developed by the PaddleOCR team, trained based on PP-OCRv3_mobile_rec and specifically optimized for Cyrillic script.

An ultra-lightweight English text line recognition model developed by the PaddleOCR team, supporting the recognition of English and digital characters

PP-OCRv5_mobile_rec is the latest generation of text line recognition model developed by the PaddleOCR team. It supports the recognition of four languages: Simplified Chinese, Traditional Chinese, English, and Japanese, and is suitable for various complex text scenarios.

A lightweight instruction fine-tuned language model optimized for mobile deployment

Depth Anything V2 is a depth estimation model based on the DPT architecture, utilizing a DINOv2 backbone network. It achieves fine and robust depth prediction through training on large-scale synthetic and real-world data.

YOLOv3 is an efficient object detection model capable of real-time localization and classification of 80 different objects in images.

MobileCLIP-B is an efficient image-text model that achieves fast inference through multimodal reinforcement training and excels in zero-shot image classification tasks.

MobileCLIP-S2 is an efficient text-image model that achieves fast zero-shot image classification through multimodal reinforcement training.

MobileCLIP-S0 is an efficient image-text model achieved through multimodal reinforcement training, significantly improving speed and size efficiency while maintaining high performance.

Depth Anything is a depth estimation model based on the DPT architecture and DINOv2 backbone network, trained on approximately 62 million images, achieving state-of-the-art results in relative and absolute depth estimation tasks.

MobileCLIP S2 is a lightweight vision-language model focused on image feature extraction and zero-shot image classification tasks.

MobileCLIP S0 is an ONNX adaptation of Apple's ml-mobileclip project, a zero-shot image classification model optimized for mobile devices.

MobileNet V2 is a lightweight convolutional neural network designed for mobile devices, achieving a balance between latency, model size, and accuracy.

MobileNet V2 is a small, low-latency, low-power vision model specifically optimized for mobile devices

This model is a fine-tuned version of the uncased MobileBERT model on the Multi-Genre Natural Language Inference (MNLI) task, suitable for zero-shot classification tasks.

MobileBERT is a lightweight variant of BERT, optimized for mobile devices while maintaining high performance.

A sentiment recognition model optimized based on MobileBERT architecture, specifically designed for the EmoContext dataset

A lightweight question-answering model based on MobileBERT, fine-tuned on the SQuAD v1.1 dataset, suitable for English Q&A tasks.