anan19990108/yolov10n_tflite

YOLOv10n TensorFlow Lite Model (Edge AI Ready)

This repository provides TensorFlow Lite versions of YOLOv10n optimized for Edge AI deployment, especially on Qualcomm Snapdragon platforms using TFLite delegates such as:

• QNN NPU
• GPU delegate
• CPU fallback

These models are suitable for:

• Android AI applications
• Embedded Linux AI pipelines
• Real-time object detection systems
• Qualcomm Edge AI platforms (QCS8550 / RB-series / Snapdragon X Elite)

Example: Android Deployment (TFLite)

Typical delegate priority:

QNN_NPU → GPU → CPU

Example initialization:

Interpreter.Options options = new Interpreter.Options();
options.setNumThreads(4);

Recommended delegates:

QNN delegate (preferred)
GPU delegate
NNAPI delegate

---

Performance Target Scenario

Designed for real-time Edge AI applications such as:

• USB camera detection
• IP camera detection
• multi-stream video analytics
• smart surveillance
• robotics perception pipelines

Validated usage scenario:

1080p video input
Real-time inference pipeline
Qualcomm NPU acceleration
Android TextureView rendering pipeline

---

Tested Platforms (Recommended)

This model is suitable for:

• Qualcomm QCS8550
• Qualcomm RB-series platforms
• Snapdragon X Elite
• Snapdragon 8 Gen series
• Android Edge AI devices